From weiyu at cs.cmu.edu  Thu Jan 18 08:25:37 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Thu, 18 Jan 2018 05:25:37 -0800
Subject: [AI Seminar] Call for Talks in AI Seminar
Message-ID: <CABzq7er00YODSZUO2m8=Z7g0xOoLFs5ZA2O2rh-POEMh16HaEg@mail.gmail.com>

Dear all,

Happy new year and welcome back to school!

The CMU AI Seminar <http://www.cs.cmu.edu/~aiseminar/> this year will start
from next Tuesday. Please sign up for contributing a talk! Again, according
to the previous experience, the slots will be filled up soon, so please do
it as soon as you can!

Looking forward to seeing you at the seminars soon!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180118/becf4790/attachment.html>

From weiyu at cs.cmu.edu  Sun Jan 21 06:04:32 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sun, 21 Jan 2018 03:04:32 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Nika Haghtalab -- Jan
	23
Message-ID: <CABzq7er-WFtKq2n65f6Lmn4AZj+q+BrKd3ZJcr1Z4erPmUzBPw@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, Jan 23, at noon in GHC 6115 for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Nika Haghtalab <https://www.cs.cmu.edu/~nhaghtal/> will give
the following talk:

Title: Machine learning by the people, for the people

Abstract:

Typical analysis of learning algorithms considers their outcome in
isolation from the effects that they may have on the process that generates
the data or the entity that is interested in learning. However, current
technological trends mean that people and organizations increasingly
interact with learning systems, making it necessary to consider these
effects, which fundamentally change the nature of learning and the
challenges involved. In this talk, I will explore three lines of research
from my work on the theoretical aspects of machine learning and algorithmic
economics that account for these interactions: learning optimal policies in
game-theoretic settings, without an accurate behavioral model, by
interacting with people; managing people's expertise and resources in
data-collection and machine learning; and collaborative learning in a
setting where multiple learners interact with each other to discover
similar underlying concepts.

Bio:
Nika Haghtalab is a Ph.D. candidate at the Computer Science Department of
Carnegie Mellon University, co-advised by Avrim Blum and Ariel Procaccia.
Her research interests include learning theory and algorithmic economics.
She is a recipient of the IBM and Microsoft Research Ph.D. fellowships and
the Siebel Scholarship.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180121/b6ee0c17/attachment.html>

From weiyu at cs.cmu.edu  Mon Jan 22 17:34:10 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 22 Jan 2018 14:34:10 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Nika Haghtalab --
	Jan 23
In-Reply-To: <CABzq7er-WFtKq2n65f6Lmn4AZj+q+BrKd3ZJcr1Z4erPmUzBPw@mail.gmail.com>
References: <CABzq7er-WFtKq2n65f6Lmn4AZj+q+BrKd3ZJcr1Z4erPmUzBPw@mail.gmail.com>
Message-ID: <CABzq7equTyr=UnBUcPNysjF21UNmgZZUzZorCyzJW0wOB+qbKw@mail.gmail.com>

A gentle reminder that the talk will happen tomorrow (Tuesday) noon at *GHC
6115*.



On Sun, Jan 21, 2018 at 3:04 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, Jan 23, at noon in GHC 6115
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Nika Haghtalab <https://www.cs.cmu.edu/~nhaghtal/> will give
> the following talk:
>
> Title: Machine learning by the people, for the people
>
> Abstract:
>
> Typical analysis of learning algorithms considers their outcome in
> isolation from the effects that they may have on the process that generates
> the data or the entity that is interested in learning. However, current
> technological trends mean that people and organizations increasingly
> interact with learning systems, making it necessary to consider these
> effects, which fundamentally change the nature of learning and the
> challenges involved. In this talk, I will explore three lines of research
> from my work on the theoretical aspects of machine learning and algorithmic
> economics that account for these interactions: learning optimal policies in
> game-theoretic settings, without an accurate behavioral model, by
> interacting with people; managing people's expertise and resources in
> data-collection and machine learning; and collaborative learning in a
> setting where multiple learners interact with each other to discover
> similar underlying concepts.
>
> Bio:
> Nika Haghtalab is a Ph.D. candidate at the Computer Science Department of
> Carnegie Mellon University, co-advised by Avrim Blum and Ariel Procaccia.
> Her research interests include learning theory and algorithmic economics.
> She is a recipient of the IBM and Microsoft Research Ph.D. fellowships and
> the Siebel Scholarship.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180122/edb590d1/attachment.html>

From weiyu at cs.cmu.edu  Sun Jan 28 00:47:20 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 27 Jan 2018 21:47:20 -0800
Subject: [AI Seminar] No AI Seminar Next Week
Message-ID: <CABzq7ep8WGzd3aNZOWzg5XbRYYE4ngRnRBn44Z5oQuhBRgaRjw@mail.gmail.com>

There won't be an AI Seminar next week. We will resume the week after next.
Good luck with the incoming massive paper deadlines!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180127/fa5ea8d6/attachment.html>

From weiyu at cs.cmu.edu  Sat Feb  3 10:28:20 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 3 Feb 2018 07:28:20 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Reihaneh Rabbany --
	Feb 06
Message-ID: <CABzq7erQp1MLMva=MGtHbYF6Vjw8Hz-SjG86y6GGNELk8EsBvA@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, Feb 06, at noon in *NSH 1507*
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Reihaneh Rabbany <http://www.reirab.com/> will give the
following talk:

Title: Mining Connections

Abstract:

Connections are ubiquitous in different domains; from collaborations and
associations between people, to proteins and genes interacting to carry out
a biological function. To study the structure and dynamics of the
interconnected world around us, we need models and algorithms that
integrate three principal elements: connection, content, and time. In order
to achieve these integrated models, I will discuss bringing machine
learning, network science, and data mining together. As a step toward this
direction, I talk about narrowing the gap between clustering (based on
content), community detection (using connections), and extending them to
the dynamic settings (time). Throughout the talk, I will give examples from
the broader impacts of this research in computational social science,
combating online human trafficking, and educational data mining.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180203/358d8766/attachment.html>

From weiyu at cs.cmu.edu  Mon Feb  5 10:39:02 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 5 Feb 2018 07:39:02 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Reihaneh Rabbany
	-- Feb 06
In-Reply-To: <CABzq7erQp1MLMva=MGtHbYF6Vjw8Hz-SjG86y6GGNELk8EsBvA@mail.gmail.com>
References: <CABzq7erQp1MLMva=MGtHbYF6Vjw8Hz-SjG86y6GGNELk8EsBvA@mail.gmail.com>
Message-ID: <CABzq7eq1ogtR2Dn4Jc7Ld3MFhcE5NuX+3ZxWPDFyaYMBdDkkyA@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in *NSH
1507.*

On Sat, Feb 3, 2018 at 7:28 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, Feb 06, at noon in *NSH 1507*
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Reihaneh Rabbany <http://www.reirab.com/> will give the
> following talk:
>
> Title: Mining Connections
>
> Abstract:
>
> Connections are ubiquitous in different domains; from collaborations and
> associations between people, to proteins and genes interacting to carry out
> a biological function. To study the structure and dynamics of the
> interconnected world around us, we need models and algorithms that
> integrate three principal elements: connection, content, and time. In order
> to achieve these integrated models, I will discuss bringing machine
> learning, network science, and data mining together. As a step toward this
> direction, I talk about narrowing the gap between clustering (based on
> content), community detection (using connections), and extending them to
> the dynamic settings (time). Throughout the talk, I will give examples from
> the broader impacts of this research in computational social science,
> combating online human trafficking, and educational data mining.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180205/7819c437/attachment.html>

From weiyu at cs.cmu.edu  Sat Feb 10 08:38:41 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 10 Feb 2018 05:38:41 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Ido Erev (Technion -
 Israel Institute of Technology) -- Feb 13
Message-ID: <CABzq7ep8G1jsUCxfxibutA1TFQTwmM3qtSNobLgZ9=G-Poi8ew@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, Feb 13, at noon in *NSH 1507* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Ido Erev
<https://web.iem.technion.ac.il/en/people/userprofile/erev.html> will give
the following talk:

Title: When and how can social scientists add value to data scientists? A
choice prediction competition for human decision making

Abstract:

Behavioral decision research highlights interesting choice anomalies, and
proposes elegant cognitive models that can explain these phenomena. Yet, it
is often easier to predict behavior with theory-free machine learning tools
than with the leading cognitive models. One reason for the difficulty in
deriving general predictions using cognitive models is that different
models are often proposed to explain different phenomena. It is then
unclear which model to use to address a new task. The current talk reviews
recent research and describes a new choice prediction competition project (
https://cpc18.wordpress.com) that tries to address this problem.

Based on research with Ori Plonsky, Reut Apel, Eyal Ert and Moshe
Tennenholtz.


Bio:

Ido Erev is the President Elect of the European Association for Decision
Making, Professor of Behavioral Science at the Technion, and Research
Environment Professor in Warwick Business School. His work focuses on the
impact of economics incentives on choice behavior. It suggests that the
initial reaction to a description of the incentive structure reflects
overweighting of rare events, but experience reverses this bias.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180210/3d2d43df/attachment.html>

From weiyu at cs.cmu.edu  Mon Feb 12 06:58:25 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 12 Feb 2018 03:58:25 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Ido Erev
 (Technion - Israel Institute of Technology) -- Feb 13
In-Reply-To: <CABzq7ep8G1jsUCxfxibutA1TFQTwmM3qtSNobLgZ9=G-Poi8ew@mail.gmail.com>
References: <CABzq7ep8G1jsUCxfxibutA1TFQTwmM3qtSNobLgZ9=G-Poi8ew@mail.gmail.com>
Message-ID: <CABzq7eo4O0OBYX-xYzNH2AT3WGd7gg_up=xnRS==pgVvwK6TBw@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in *NSH
1507.*

On Sat, Feb 10, 2018 at 5:38 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, Feb 13, at noon in *NSH 1507* for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Ido Erev
> <https://web.iem.technion.ac.il/en/people/userprofile/erev.html> will
> give the following talk:
>
> Title: When and how can social scientists add value to data scientists? A
> choice prediction competition for human decision making
>
> Abstract:
>
> Behavioral decision research highlights interesting choice anomalies, and
> proposes elegant cognitive models that can explain these phenomena. Yet, it
> is often easier to predict behavior with theory-free machine learning tools
> than with the leading cognitive models. One reason for the difficulty in
> deriving general predictions using cognitive models is that different
> models are often proposed to explain different phenomena. It is then
> unclear which model to use to address a new task. The current talk reviews
> recent research and describes a new choice prediction competition project (
> https://cpc18.wordpress.com) that tries to address this problem.
>
> Based on research with Ori Plonsky, Reut Apel, Eyal Ert and Moshe
> Tennenholtz.
>
>
> Bio:
>
> Ido Erev is the President Elect of the European Association for Decision
> Making, Professor of Behavioral Science at the Technion, and Research
> Environment Professor in Warwick Business School. His work focuses on the
> impact of economics incentives on choice behavior. It suggests that the
> initial reaction to a description of the incentive structure reflects
> overweighting of rare events, but experience reverses this bias.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180212/a4a2d966/attachment.html>

From weiyu at cs.cmu.edu  Sat Feb 17 15:40:38 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 17 Feb 2018 12:40:38 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Alec Koppel -- Feb 20
Message-ID: <CABzq7eo__30aNeTumkpsaDjwgfO7wVGyO3-R3UcwjeccCnFhoA@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, Feb 20, at noon in *NSH 1507* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Alec Koppel <http://koppel.bitballoon.com/> will give the
following talk:

Title: Nonparametric Stochastic Methods for Continuous Reinforcement
Learning

Abstract:

Reinforcement learning is a generic framework to describe an autonomous
agent seeking to learn behavior sequentially in uncertain environments
based on rewards. This framework has gained increasing relevance for
autonomous control, management science, and econometrics. Unfortunately,
heuristics or intractably complicated tools are still prevalent when state
and action spaces are continuous. In this talk, we develop new algorithms
for estimating the value function or action-value function in continuous
Markov Decision Problems (MDPs). The core of these methods are
nonparametric (kernelized) extensions of stochastic quasi-gradient methods
operating in tandem with sparse subspace projections. The resulting tools
yield the first convergence results for Value or Q-function estimation when
these functions have an infinite nonlinear parameterization, addressing in
the affirmative a long-standing open question posed by Tsistiklis and Van
Roy (1997). We then demonstrate on the classic Mountain Car domain that we
can obtain comparable performance to existing approaches to TD or Q
learning with orders of magnitude fewer data samples and interpretable
representations of the learned functions.

Biography:

Alec Koppel began as a Research Scientist at the U.S. Army Research
Laboratory in the Computational and Information Sciences Directorate in
September of 2017. He completed his master's degree in Statistics and
doctorate in Electrical and Systems Engineering, both at the University of
Pennsylvania (Penn) in August of 2017. He is also a participant in the
Science, Mathematics, and Research for Transformation (SMART) Scholarship
Program sponsored by the American Society of Engineering Education. Before
coming to Penn, he completed his Master's degree in Systems Science and
Mathematics and Bachelor's Degree in Mathematics, both at Washington
University in St. Louis (WashU), Missouri. His research interests are in
the areas of signal processing, optimization and learning theory. His
current work focuses on optimization and learning methods for streaming
data applications, with an emphasis on problems arising in autonomous
systems. He co-authored a paper selected as a Best Paper Finalist at the
2017 IEEE Asilomar Conference on Signals, Systems, and Computers.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180217/5cca1527/attachment.html>

From weiyu at cs.cmu.edu  Mon Feb 19 20:11:33 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 19 Feb 2018 17:11:33 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Alec Koppel --
	Feb 20
In-Reply-To: <CABzq7eo__30aNeTumkpsaDjwgfO7wVGyO3-R3UcwjeccCnFhoA@mail.gmail.com>
References: <CABzq7eo__30aNeTumkpsaDjwgfO7wVGyO3-R3UcwjeccCnFhoA@mail.gmail.com>
Message-ID: <CABzq7eoRzWgLZcmuCH_9VJ9OtM1EtkkpYnuUcOimJ_gAusVWkg@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in *NSH
1507*.

On Sat, Feb 17, 2018 at 12:40 PM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, Feb 20, at noon in *NSH 1507* for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Alec Koppel <http://koppel.bitballoon.com/> will give the
> following talk:
>
> Title: Nonparametric Stochastic Methods for Continuous Reinforcement
> Learning
>
> Abstract:
>
> Reinforcement learning is a generic framework to describe an autonomous
> agent seeking to learn behavior sequentially in uncertain environments
> based on rewards. This framework has gained increasing relevance for
> autonomous control, management science, and econometrics. Unfortunately,
> heuristics or intractably complicated tools are still prevalent when state
> and action spaces are continuous. In this talk, we develop new algorithms
> for estimating the value function or action-value function in continuous
> Markov Decision Problems (MDPs). The core of these methods are
> nonparametric (kernelized) extensions of stochastic quasi-gradient methods
> operating in tandem with sparse subspace projections. The resulting tools
> yield the first convergence results for Value or Q-function estimation when
> these functions have an infinite nonlinear parameterization, addressing in
> the affirmative a long-standing open question posed by Tsistiklis and Van
> Roy (1997). We then demonstrate on the classic Mountain Car domain that we
> can obtain comparable performance to existing approaches to TD or Q
> learning with orders of magnitude fewer data samples and interpretable
> representations of the learned functions.
>
> Biography:
>
> Alec Koppel began as a Research Scientist at the U.S. Army Research
> Laboratory in the Computational and Information Sciences Directorate in
> September of 2017. He completed his master's degree in Statistics and
> doctorate in Electrical and Systems Engineering, both at the University of
> Pennsylvania (Penn) in August of 2017. He is also a participant in the
> Science, Mathematics, and Research for Transformation (SMART) Scholarship
> Program sponsored by the American Society of Engineering Education. Before
> coming to Penn, he completed his Master's degree in Systems Science and
> Mathematics and Bachelor's Degree in Mathematics, both at Washington
> University in St. Louis (WashU), Missouri. His research interests are in
> the areas of signal processing, optimization and learning theory. His
> current work focuses on optimization and learning methods for streaming
> data applications, with an emphasis on problems arising in autonomous
> systems. He co-authored a paper selected as a Best Paper Finalist at the
> 2017 IEEE Asilomar Conference on Signals, Systems, and Computers.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180219/6dfbe271/attachment.html>

From cga at cs.cmu.edu  Fri Feb 23 00:28:36 2018
From: cga at cs.cmu.edu (Chris Atkeson)
Date: Fri, 23 Feb 2018 00:28:36 -0500 (EST)
Subject: [AI Seminar] Talk: Learning deep representations for perception,
 reasoning, and decision-making
Message-ID: <Pine.LNX.4.63.1802230019300.21860@score.GRAPHICS.cs.cmu.edu>

Learning deep representations for perception, reasoning, and decision-making
Honglak Lee
Monday, Feb 26, 10:00am, NSH 3305

Abstract:
Over the recent years, deep learning has emerged as a powerful method for 
learning feature representations from complex input data, and it has been 
greatly successful in many domains, such as computer vision, speech 
recognition, and language processing. While many deep learning algorithms 
focus on standard discriminative tasks with explicit supervision (e.g., 
classification), we aim for learning deep representations with less 
supervision that can serve as explanatory models of data and allow for 
richer inference, reasoning, and control.

First, I will present techniques for learning deep representations in 
weakly-supervised settings that disentangle underlying factors of 
variation. These learned representations model intricate interaction 
between underlying factors of variations in the data (e.g., pose and 
morphology for 3d objects in images) and allows for hypothetical 
reasoning and improved control (e.g., grasping of objects). I will also 
talk about learning via analogy-making and its connection to 
disentangling. In the second part of the talk, I will describe my work on 
learning deep representations from multiple heterogeneous input 
modalities. Specifically, I will present a multimodal learning framework 
via conditional prediction that explicitly encourages cross-modal 
associations.  This framework provides a theoretical guarantee about 
learning a joint distribution and explains recent progress in deep 
architectures that interface vision and language, such as caption 
generation and conditional image synthesis. I will also describe related 
ongoing work on learning joint embedding from images and text for 
zero-shot learning. Finally, I will present ongoing work on integrating 
deep learning and reinforcement learning. Specifically, I will talk about 
how learning a predictive generative model from  sequential input data can 
be useful for reinforcement learning. I also will talk about a 
memory-based architecture that helps sequential decision making in 
a first-person view and active perception setting, as well as zero-shot 
multi-task generalization with hierarchical reinforcement learning given 
task descriptions.

Bio:
Honglak Lee is a Research Scientist at Google Brain and an Associate 
Professor of Computer Science and Engineering at the University of 
Michigan, Ann Arbor. He received his Ph.D. from Computer Science 
Department at Stanford University in 2010, advised by Prof. Andrew Ng.
His research focuses on deep learning and representation learning, which 
spans over unsupervised and supervised and semi-supervised learning, 
transfer learning, structured prediction, reinforcement learning, and 
optimization. His methods have been successfully applied to computer 
vision and other perception problems. He received best paper awards at 
ICML 2009 and CEAS 2005. He has served as area chairs of ICML, NIPS, 
ICLR, ICCV, CVPR, ECCV, AAAI, and IJCAI, as well as a guest editor of 
IEEE TPAMI Special Issue on Learning Deep Architectures, an editorial 
board member of Neural Networks, and an associate editor of IEEE TPAMI. 
He received the Google Faculty Research Award (2011), NSF CAREER Award 
(2015), and was selected as one of AI's 10 to Watch by IEEE Intelligent 
Systems (2013) and a research fellow by Alfred P. Sloan Foundation 
(2016).

From weiyu at cs.cmu.edu  Sat Feb 24 07:04:17 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 24 Feb 2018 04:04:17 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Ariel Procaccia --
	Feb 27
Message-ID: <CABzq7erGEJzgOuQHgBHuQPbAVu+FEpi+zHbVBALkpCtJaYsRwg@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, Feb 27, at noon in *NSH 1507* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday,  Ariel Procaccia <http://procaccia.info/> will give the
following talk:

Title: Extreme Democracy

Abstract:

Technological advances have changed every aspect of our lives in recent
decades, yet, for the most part, the same systems of democratic decision
making have been in place for centuries. I will argue that computer
scientists can help rethink the practice of democracy, as well as its
potential applications. I will focus on three emerging paradigms that go
far beyond your run-of-the-mill election: (i) liquid democracy, an approach
that allows voters to transitively delegate their votes; (ii) participatory
budgeting, whereby residents collectively decide how to spend their local
government's budget; and (iii) virtual democracy, which employs instant
elections among machine learning models of real voters to address the grand
AI challenge of ethical decision making.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180224/c14c14cc/attachment.html>

From weiyu at cs.cmu.edu  Mon Feb 26 07:33:13 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 26 Feb 2018 04:33:13 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Ariel Procaccia
	-- Feb 27
In-Reply-To: <CABzq7erGEJzgOuQHgBHuQPbAVu+FEpi+zHbVBALkpCtJaYsRwg@mail.gmail.com>
References: <CABzq7erGEJzgOuQHgBHuQPbAVu+FEpi+zHbVBALkpCtJaYsRwg@mail.gmail.com>
Message-ID: <CABzq7eq7mFRbAN0WsS=suE08fhLNG5HZCOOW601XKruj0Pe1PQ@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon at *NSH
1507*.

On Sat, Feb 24, 2018 at 4:04 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, Feb 27, at noon in *NSH 1507* for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday,  Ariel Procaccia <http://procaccia.info/> will give the
> following talk:
>
> Title: Extreme Democracy
>
> Abstract:
>
> Technological advances have changed every aspect of our lives in recent
> decades, yet, for the most part, the same systems of democratic decision
> making have been in place for centuries. I will argue that computer
> scientists can help rethink the practice of democracy, as well as its
> potential applications. I will focus on three emerging paradigms that go
> far beyond your run-of-the-mill election: (i) liquid democracy, an approach
> that allows voters to transitively delegate their votes; (ii) participatory
> budgeting, whereby residents collectively decide how to spend their local
> government's budget; and (iii) virtual democracy, which employs instant
> elections among machine learning models of real voters to address the grand
> AI challenge of ethical decision making.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180226/164cf993/attachment.html>

From cga at cs.cmu.edu  Sun Feb 25 15:01:13 2018
From: cga at cs.cmu.edu (Chris Atkeson)
Date: Sun, 25 Feb 2018 15:01:13 -0500 (EST)
Subject: [AI Seminar] Talk: Computational Sensorimotor Learning
In-Reply-To: <Pine.LNX.4.63.1802230019300.21860@score.GRAPHICS.cs.cmu.edu>
References: <Pine.LNX.4.63.1802230019300.21860@score.GRAPHICS.cs.cmu.edu>
Message-ID: <Pine.LNX.4.63.1802251456400.6820@score.GRAPHICS.cs.cmu.edu>

In addition to the talk on Monday morning,
> Learning deep representations for perception, reasoning, and decision-making
> Honglak Lee
> Monday, Feb 26, 10:00am, NSH 3305

we have an exciting talk on Tuesday morning:

SCS Faculty Candidate:???Pulkit Agrawal
Tuesday, February 27th 10:00am NSH3305

Host: Ryan Tibshirani

RI/CSD/ML

Computational Sensorimotor Learning

Abstract: An open question in artificial intelligence is how to endow
agents with common sense knowledge that humans naturally seem to
possess. A prominent theory in child development posits that human
infants gradually acquire such knowledge by the process of
experimentation. According to this theory, even the seemingly
frivolous play of infants is a manifestation of experiments conducted
by them to learn about their environment.???Inspired by this view of
biological sensorimotor learning, I will present my work on building
artificial agents that use the paradigm of experimentation to explore
and condense their experience into models that enable them to solve
new problems. I will discuss the effectiveness of my approach and
open issues using case studies of a robot learning to push objects,
manipulate ropes, finding its way in office environments and an agent
learning to play video games merely based on the incentive of
conducting experiments.

Bio:???Pulkit???is a Ph.D. Student in the department of computer science
at UC Berkeley and Co-Founder of SafelyYou Inc. He is advised by Dr.
Jitendra Malik and his research spans robotics, deep learning,
computer vision and computational neuroscience.???Pulkit???completed his
bachelors in Electrical Engineering from IIT Kanpur and was awarded
the Director's???Gold???Medal.???His work has appeared multiple times in
MIT???Tech???Review, Quanta, New Scientist, NYPost etc.???He is a 
recipient of Signatures Fellow Award, Fulbright Science and Technology 
Award, Goldman Sachs Global Leadership Award, OPJEMS, Sridhar Memorial 
Prize and IIT Kanpur's Academic Excellence Awards among
others.???Pulkit???holds a "Sangeet Prabhakar" (equivalent to bachelors
in Indian classical???music) and occasionally performs in
music???concerts.???

From weiyu at cs.cmu.edu  Sun Mar  4 07:30:04 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sun, 4 Mar 2018 04:30:04 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Chenyan Xiong --
	March 06
Message-ID: <CABzq7eoyMK49r5UHgEqZ45safNokDx7GriTRR7_mJh=K6tdqMA@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, March 06, at noon in *NSH 1507* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday,  Chenyan Xiong <https://www.cs.cmu.edu/~cx/> will give the
following talk:

Title: Text Representation, Retrieval, and Understanding with Knowledge
Graphs

Abstract:

Search engines and other information systems have started to evolve from
retrieving documents to providing more intelligent information access.
However, the evolution is still in its infancy due to computers' limited
ability in representing and understanding human language. This talk will
present my work addressing these challenges with knowledge graphs. The
first part is about utilizing entities from knowledge graphs to improve
search. I will discuss how we build better text representations with
entities and how the entity-based text representations improve text
retrieval. The second part is about better text understanding through
modeling entity salience (importance), as well as how the improved text
understanding helps search under both feature-based and neural ranking
settings. This talk concludes with future directions towards the next
generation of intelligent information systems.


Bio:

Chenyan Xiong is a Ph.D. candidate at Carnegie Mellon University. His
research lies in the intersection of machine learning and information
retrieval. His current research focus is on improving text representation
and understanding in real-world information systems using knowledge graphs
and neural networks. He is a recipient of Allen Institute for Artificial
Intelligence research fellowship. Besides publishing papers, he also
co-organizes NTCIR WWW Tracks about deep learning for search, the first
SIGIR workshop on knowledge graphs and semantics for text retrieval and
analysis, and a special issue in Information Retrieval Journal about
knowledge graph for IR.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180304/b58e9d6b/attachment.html>

From weiyu at cs.cmu.edu  Mon Mar  5 09:02:27 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 5 Mar 2018 06:02:27 -0800
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Chenyan Xiong --
	March 06
In-Reply-To: <CABzq7eoyMK49r5UHgEqZ45safNokDx7GriTRR7_mJh=K6tdqMA@mail.gmail.com>
References: <CABzq7eoyMK49r5UHgEqZ45safNokDx7GriTRR7_mJh=K6tdqMA@mail.gmail.com>
Message-ID: <CABzq7eqxPaCbngy_KQ+tJUYu02-Fhg2cmKmkrG7EvNCohEwYrA@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in *NSH
1507*

On Sun, Mar 4, 2018 at 4:30 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, March 06, at noon in *NSH
> 1507* for AI Seminar sponsored by Apple. To learn more about the seminar
> series, please visit the AI Seminar webpage
> <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday,  Chenyan Xiong <https://www.cs.cmu.edu/~cx/> will give the
> following talk:
>
> Title: Text Representation, Retrieval, and Understanding with Knowledge
> Graphs
>
> Abstract:
>
> Search engines and other information systems have started to evolve from
> retrieving documents to providing more intelligent information access.
> However, the evolution is still in its infancy due to computers' limited
> ability in representing and understanding human language. This talk will
> present my work addressing these challenges with knowledge graphs. The
> first part is about utilizing entities from knowledge graphs to improve
> search. I will discuss how we build better text representations with
> entities and how the entity-based text representations improve text
> retrieval. The second part is about better text understanding through
> modeling entity salience (importance), as well as how the improved text
> understanding helps search under both feature-based and neural ranking
> settings. This talk concludes with future directions towards the next
> generation of intelligent information systems.
>
>
> Bio:
>
> Chenyan Xiong is a Ph.D. candidate at Carnegie Mellon University. His
> research lies in the intersection of machine learning and information
> retrieval. His current research focus is on improving text representation
> and understanding in real-world information systems using knowledge graphs
> and neural networks. He is a recipient of Allen Institute for Artificial
> Intelligence research fellowship. Besides publishing papers, he also
> co-organizes NTCIR WWW Tracks about deep learning for search, the first
> SIGIR workshop on knowledge graphs and semantics for text retrieval and
> analysis, and a special issue in Information Retrieval Journal about
> knowledge graph for IR.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180305/01d4bed9/attachment.html>

From weiyu at cs.cmu.edu  Sat Mar 10 03:35:40 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 10 Mar 2018 00:35:40 -0800
Subject: [AI Seminar] No AI Seminar Next Week
Message-ID: <CABzq7eqfFe2u3-J5uUUZQBPs_vuU92q0BLvFP5S7oq_dkyVAXw@mail.gmail.com>

There will be no AI Seminar next week. Happy spring break, everyone!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180310/8b5b924d/attachment.html>

From arielpro at cs.cmu.edu  Tue Mar 13 16:14:11 2018
From: arielpro at cs.cmu.edu (Ariel Procaccia)
Date: Tue, 13 Mar 2018 16:14:11 -0400
Subject: [AI Seminar] Update: Talk by Omer Reingold on March 29
Message-ID: <CADqQPTMh09HT13Usp4L=1qELKfyKAODBF32a42ba6Y5jBiUKuQ@mail.gmail.com>

Catherine Copetas tells me that the location has changed to Hamerschlag
Hall 1107.

Cheers,
Ariel

On Tue, Mar 13, 2018 at 4:04 PM, Ariel Procaccia <arielpro at cs.cmu.edu>
wrote:

> Of possible interest:
>
> *ECE Seminar: *Calibration for the (computationally-identifiable) masses
>
> *Starts at: *March 29, 2018 4:30 PM
>
> *Ends at: *6:00 PM
>
> *Location: *DH A302
>
> *Speaker: *Dr. Omer Reingold
>
> *Affiliation: *Stanford University
>
> *Refreshments provided: *Yes
>
> Link to Abstract
> <http://www.ece.cmu.edu/news/calendar/2018/03/ece-grad-seminar-reingold.pdf>
>
>   iCalendar
> <http://www.ece.cmu.edu/news/calendar/2018/03/ece-grad-seminar-reingold.ics>
>
> *Details:*
>
> Abstract:
> As algorithms increasingly inform and influence decisions made about
> individuals, it becomes increasingly important to address concerns that
> these algorithms might be discriminatory. The output of an algorithm can be
> discriminatory for many reasons, most notably: (1) the data used to train
> the algorithm might be biased (in various ways) to favor certain
> populations over others; (2) the analysis of this training data might
> inadvertently or maliciously introduce biases that are not borne out in the
> data. This work focuses on the latter concern.
>
> We develop and study multicalbration as a new measure of algorithmic
> fairness that aims to mitigate concerns about discrimination that is
> introduced in the process of learning a predictor from data.
> Multicalibration guarantees accurate (calibrated) predictions for every
> subpopulation that can be identified within a specified class of
> computations. We think of the class as being quite rich, in particular it
> can contain many and overlapping subgroups of a protected group.
>
> We show that in many settings this strong notion of protection from
> discrimination is both attainable and aligned with the goal of obtaining
> accurate predictions. Along the way, we present new algorithms for learning
> a multicalibrated predictor, study the computational complexity of this
> task, and draw new connections to computational learning models such as
> agnostic learning.
>
> Joint work with Ursula Hebert-Johnson, Michael P. Kim and Guy Rothblum
>
> Bio:
> Omer Reingold is a Professor of Computer Science at Stanford University.
> Past positions include Samsung Research America, the Weizmann Institute of
> Science, Microsoft Research, the Institute for Advanced Study in Princeton,
> NJ and AT&T Labs. His research is in the Foundations of Computer Science
> and most notably in Computational Complexity and the Foundations of
> Cryptography with emphasis on randomness, derandomization and explicit
> combinatorial constructions. He has a keen interest in the societal impact
> of computation. He is an ACM Fellow and among his distinctions are the 2005
> Grace Murray Hopper Award and the 2009 G?del Prize.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180313/9dcbf177/attachment.html>

From arielpro at cs.cmu.edu  Tue Mar 13 16:04:35 2018
From: arielpro at cs.cmu.edu (Ariel Procaccia)
Date: Tue, 13 Mar 2018 16:04:35 -0400
Subject: [AI Seminar] Talk by Omer Reingold on March 29
Message-ID: <CADqQPTN8B9GKwRCOsyxqLXL45W4n-bF+8DntD5pouHkKgoTkhg@mail.gmail.com>

Of possible interest:

*ECE Seminar: *Calibration for the (computationally-identifiable) masses

*Starts at: *March 29, 2018 4:30 PM

*Ends at: *6:00 PM

*Location: *DH A302

*Speaker: *Dr. Omer Reingold

*Affiliation: *Stanford University

*Refreshments provided: *Yes

Link to Abstract
<http://www.ece.cmu.edu/news/calendar/2018/03/ece-grad-seminar-reingold.pdf>

  iCalendar
<http://www.ece.cmu.edu/news/calendar/2018/03/ece-grad-seminar-reingold.ics>

*Details:*

Abstract:
As algorithms increasingly inform and influence decisions made about
individuals, it becomes increasingly important to address concerns that
these algorithms might be discriminatory. The output of an algorithm can be
discriminatory for many reasons, most notably: (1) the data used to train
the algorithm might be biased (in various ways) to favor certain
populations over others; (2) the analysis of this training data might
inadvertently or maliciously introduce biases that are not borne out in the
data. This work focuses on the latter concern.

We develop and study multicalbration as a new measure of algorithmic
fairness that aims to mitigate concerns about discrimination that is
introduced in the process of learning a predictor from data.
Multicalibration guarantees accurate (calibrated) predictions for every
subpopulation that can be identified within a specified class of
computations. We think of the class as being quite rich, in particular it
can contain many and overlapping subgroups of a protected group.

We show that in many settings this strong notion of protection from
discrimination is both attainable and aligned with the goal of obtaining
accurate predictions. Along the way, we present new algorithms for learning
a multicalibrated predictor, study the computational complexity of this
task, and draw new connections to computational learning models such as
agnostic learning.

Joint work with Ursula Hebert-Johnson, Michael P. Kim and Guy Rothblum

Bio:
Omer Reingold is a Professor of Computer Science at Stanford University.
Past positions include Samsung Research America, the Weizmann Institute of
Science, Microsoft Research, the Institute for Advanced Study in Princeton,
NJ and AT&T Labs. His research is in the Foundations of Computer Science
and most notably in Computational Complexity and the Foundations of
Cryptography with emphasis on randomness, derandomization and explicit
combinatorial constructions. He has a keen interest in the societal impact
of computation. He is an ACM Fellow and among his distinctions are the 2005
Grace Murray Hopper Award and the 2009 G?del Prize.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180313/4581815d/attachment-0001.html>

From weiyu at cs.cmu.edu  Sun Mar 18 06:22:32 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sun, 18 Mar 2018 03:22:32 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- John Dickerson
 (University of Maryland) -- March 20
Message-ID: <CABzq7erZzceL1R2RqUEMAy-X1r-GUHYF3xaif56S0GJhffd4jA@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, March 20, at noon in *NSH 1507* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday,  John Dickerson <http://jpdickerson.com/> (University of
Maryland) will give the following talk:

Title:  Diversity in Matching Markets

Abstract:

In bipartite matching problems, vertices on one side of a bipartite graph
are paired with those on the other.  In its offline variant, both sides of
the graph are known a priori; in its online variant, one side of the graph
is available offline, while vertices on the other arrive online and are
irrevocably and immediately matched (or ignored) by an algorithm.  Examples
of such problems include matching workers to firms, advertisers to
keywords, organs to patients, and riders to rideshare drivers.  Much of the
literature focuses on maximizing the total relevance---modeled via total
weight---of the matching. However, in many real-world problems, it is also
important to consider contributions of diversity: hiring a diverse pool of
candidates, displaying a relevant but diverse set of ads, and so on.

In this talk, we model the promotion of diversity in matching markets
via maximization
of a submodular function over the set of matched edges.  We present new
results in a generalization of traditional offline matching, *b*-matching,
where vertices have both lower and upper bounds on the number of adjacent
matched edges.  We also present new theoretical results in *online* submodular
bipartite matching.  Finally, we conclude with ongoing work that approaches
the problem of hiring a diverse cohort of workers through the lens of
combinatorial pure exploration (CPE) in the multiarmed bandit setting, and
discuss an ongoing experiment in this space at a large research university.

*This talk will cover joint work with Faez Ahmed, Samsara Counts, Jeff
Foster, Mark Fuge, Karthik A. Sankararaman, Candice Schumann, Aravind
Srinivasan, and Pan Xu. *


Bio:

John P Dickerson is an Assistant Professor of Computer Science at the
University of Maryland. His research centers on solving practical economic
problems using techniques from computer science, stochastic optimization,
and machine learning. He has worked extensively on theoretical and
empirical approaches to designing markets for organ allocation, dating,
admissions, and computational advertising.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180318/e467f383/attachment.html>

From weiyu at cs.cmu.edu  Tue Mar 20 01:16:49 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Tue, 20 Mar 2018 01:16:49 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- John Dickerson
 (University of Maryland) -- March 20
In-Reply-To: <CABzq7erZzceL1R2RqUEMAy-X1r-GUHYF3xaif56S0GJhffd4jA@mail.gmail.com>
References: <CABzq7erZzceL1R2RqUEMAy-X1r-GUHYF3xaif56S0GJhffd4jA@mail.gmail.com>
Message-ID: <CABzq7er+bvOk=kEs5c-LvkQwtUq4RSeF1US-t+JHKaEXxwm4cw@mail.gmail.com>

A gentle reminder that the talk will be today (Tuesday) noon at *NSH 1507*.

On Sun, Mar 18, 2018 at 6:22 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, March 20, at noon in *NSH
> 1507* for AI Seminar sponsored by Apple. To learn more about the seminar
> series, please visit the AI Seminar webpage
> <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday,  John Dickerson <http://jpdickerson.com/> (University of
> Maryland) will give the following talk:
>
> Title:  Diversity in Matching Markets
>
> Abstract:
>
> In bipartite matching problems, vertices on one side of a bipartite graph
> are paired with those on the other.  In its offline variant, both sides of
> the graph are known a priori; in its online variant, one side of the graph
> is available offline, while vertices on the other arrive online and are
> irrevocably and immediately matched (or ignored) by an algorithm.  Examples
> of such problems include matching workers to firms, advertisers to
> keywords, organs to patients, and riders to rideshare drivers.  Much of the
> literature focuses on maximizing the total relevance---modeled via total
> weight---of the matching. However, in many real-world problems, it is also
> important to consider contributions of diversity: hiring a diverse pool of
> candidates, displaying a relevant but diverse set of ads, and so on.
>
> In this talk, we model the promotion of diversity in matching markets via maximization
> of a submodular function over the set of matched edges.  We present new
> results in a generalization of traditional offline matching, *b*-matching,
> where vertices have both lower and upper bounds on the number of adjacent
> matched edges.  We also present new theoretical results in *online* submodular
> bipartite matching.  Finally, we conclude with ongoing work that approaches
> the problem of hiring a diverse cohort of workers through the lens of
> combinatorial pure exploration (CPE) in the multiarmed bandit setting, and
> discuss an ongoing experiment in this space at a large research university.
>
> *This talk will cover joint work with Faez Ahmed, Samsara Counts, Jeff
> Foster, Mark Fuge, Karthik A. Sankararaman, Candice Schumann, Aravind
> Srinivasan, and Pan Xu. *
>
>
> Bio:
>
> John P Dickerson is an Assistant Professor of Computer Science at the
> University of Maryland. His research centers on solving practical economic
> problems using techniques from computer science, stochastic optimization,
> and machine learning. He has worked extensively on theoretical and
> empirical approaches to designing markets for organ allocation, dating,
> admissions, and computational advertising.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180320/822599e7/attachment.html>

From weiyu at cs.cmu.edu  Sat Mar 24 20:50:04 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 24 Mar 2018 20:50:04 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Simon Du -- March 27
Message-ID: <CABzq7eoNCwMyOGFoB6xwZsdueVt3icPJuex+O0h5vMHjHVA7sw@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, March 27, at noon in *GHC 6115
(unusual place) *for AI Seminar sponsored by Apple. To learn more about the
seminar series, please visit the AI Seminar webpage
<http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday,  Simon Du <http://www.cs.cmu.edu/~ssdu/> will give the
following talk:

Title:  On the Power of Randomly Initialized Gradient Descent for Learning
Convolutional Neural Networks

Abstract:

Convolutional neural networks trained by randomly initialized (stochastic)
gradient descent have achieved the state-of-art performances in many
applications. However, its theoretical properties remain elusive from an
optimization point of view. In this talk, I will present two results on
explaining the success of gradient descent.

In the first part, I will show under certain structural conditions of the
input distribution, randomly initialized gradient descent provably learns a
convolutional filter with ReLU activation and average pooling. This is the
first recovery guarantee of gradient-based algorithms for learning a
convolutional filter on general input distributions.

In the second part of the talk, I will show if the input distribution is
Gaussian, then randomly initialized gradient descent with
weight-normalization learns a ReLU activated one-hidden-layer convolutional
neural network where both the convolutional weights and the output weights
are to be optimized. To the best our knowledge, this is the first recovery
guarantee of randomly initialized gradient-based algorithms for neural
networks that contain more than one layers to be learned.

This talk is based on works with Jason D. Lee, Barnabas Poczos, Aarti Singh
and Yuandong Tian.


Bio:
Simon Shaolei Du is a PhD student in the Machine Learning Department at the
School of Computer Science, Carnegie Mellon University, advised by
Professor Aarti Singh and Professor Barnabas Poczos. His research interests
broadly include topics in theoretical machine learning and statistics, such
as deep learning, matrix factorization, convex/non-convex optimization,
transfer learning, reinforcement learning, non-parametric statistics and
robust statistics. Currently he is also developing methods for precision
agriculture. In 2011, he earned his high school degree from The
Experimental High School Attached to Beijing Normal University. In 2015, he
obtained his B.S. in Engineering Math & Statistics and B.S. in Electrical
Engineering & Computer Science from University of California, Berkeley. He
has also spent time working at research labs of Microsoft and Facebook.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180324/599508dd/attachment.html>

From weiyu at cs.cmu.edu  Mon Mar 26 18:58:18 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 26 Mar 2018 18:58:18 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Simon Du -- March
	27
In-Reply-To: <CABzq7eoNCwMyOGFoB6xwZsdueVt3icPJuex+O0h5vMHjHVA7sw@mail.gmail.com>
References: <CABzq7eoNCwMyOGFoB6xwZsdueVt3icPJuex+O0h5vMHjHVA7sw@mail.gmail.com>
Message-ID: <CABzq7er=JFA1td4QnfcYv5MZxgQr7xmy1HKYyM-7d4J5tUJMng@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon at *GHC
6115 (unusual room).*

On Sat, Mar 24, 2018 at 8:50 PM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, March 27, at noon in *GHC
> 6115 (unusual place) *for AI Seminar sponsored by Apple. To learn more
> about the seminar series, please visit the AI Seminar webpage
> <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday,  Simon Du <http://www.cs.cmu.edu/~ssdu/> will give the
> following talk:
>
> Title:  On the Power of Randomly Initialized Gradient Descent for
> Learning Convolutional Neural Networks
>
> Abstract:
>
> Convolutional neural networks trained by randomly initialized (stochastic)
> gradient descent have achieved the state-of-art performances in many
> applications. However, its theoretical properties remain elusive from an
> optimization point of view. In this talk, I will present two results on
> explaining the success of gradient descent.
>
> In the first part, I will show under certain structural conditions of the
> input distribution, randomly initialized gradient descent provably learns a
> convolutional filter with ReLU activation and average pooling. This is the
> first recovery guarantee of gradient-based algorithms for learning a
> convolutional filter on general input distributions.
>
> In the second part of the talk, I will show if the input distribution is
> Gaussian, then randomly initialized gradient descent with
> weight-normalization learns a ReLU activated one-hidden-layer convolutional
> neural network where both the convolutional weights and the output weights
> are to be optimized. To the best our knowledge, this is the first recovery
> guarantee of randomly initialized gradient-based algorithms for neural
> networks that contain more than one layers to be learned.
>
> This talk is based on works with Jason D. Lee, Barnabas Poczos, Aarti
> Singh and Yuandong Tian.
>
>
> Bio:
> Simon Shaolei Du is a PhD student in the Machine Learning Department at
> the School of Computer Science, Carnegie Mellon University, advised by
> Professor Aarti Singh and Professor Barnabas Poczos. His research interests
> broadly include topics in theoretical machine learning and statistics, such
> as deep learning, matrix factorization, convex/non-convex optimization,
> transfer learning, reinforcement learning, non-parametric statistics and
> robust statistics. Currently he is also developing methods for precision
> agriculture. In 2011, he earned his high school degree from The
> Experimental High School Attached to Beijing Normal University. In 2015, he
> obtained his B.S. in Engineering Math & Statistics and B.S. in Electrical
> Engineering & Computer Science from University of California, Berkeley. He
> has also spent time working at research labs of Microsoft and Facebook.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180326/b4d2bbfd/attachment.html>

From arielpro at cs.cmu.edu  Tue Mar 27 16:59:18 2018
From: arielpro at cs.cmu.edu (Ariel Procaccia)
Date: Tue, 27 Mar 2018 16:59:18 -0400
Subject: [AI Seminar] Update: Talk by Omer Reingold on March 29
Message-ID: <CADqQPTPD1EttxE6+U5af1JdMi5JZ-U2eQjtxXWKqV_VeMxGJSw@mail.gmail.com>

Reminder: this talk is on Thursday.


> On Tue, Mar 13, 2018 at 4:04 PM, Ariel Procaccia <arielpro at cs.cmu.edu>
> wrote:
>
>> Of possible interest:
>>
>> *ECE Seminar: *Calibration for the (computationally-identifiable) masses
>>
>> *Starts at: *March 29, 2018 4:30 PM
>>
>> *Ends at: *6:00 PM
>>
>> *Location: * Hamerschlag Hall 1107
>>
>> *Speaker: *Dr. Omer Reingold
>>
>> *Affiliation: *Stanford University
>>
>> *Refreshments provided: *Yes
>>
>> Link to Abstract
>> <http://www.ece.cmu.edu/news/calendar/2018/03/ece-grad-seminar-reingold.pdf>
>>
>>   iCalendar
>> <http://www.ece.cmu.edu/news/calendar/2018/03/ece-grad-seminar-reingold.ics>
>>
>> *Details:*
>>
>> Abstract:
>> As algorithms increasingly inform and influence decisions made about
>> individuals, it becomes increasingly important to address concerns that
>> these algorithms might be discriminatory. The output of an algorithm can be
>> discriminatory for many reasons, most notably: (1) the data used to train
>> the algorithm might be biased (in various ways) to favor certain
>> populations over others; (2) the analysis of this training data might
>> inadvertently or maliciously introduce biases that are not borne out in the
>> data. This work focuses on the latter concern.
>>
>> We develop and study multicalbration as a new measure of algorithmic
>> fairness that aims to mitigate concerns about discrimination that is
>> introduced in the process of learning a predictor from data.
>> Multicalibration guarantees accurate (calibrated) predictions for every
>> subpopulation that can be identified within a specified class of
>> computations. We think of the class as being quite rich, in particular it
>> can contain many and overlapping subgroups of a protected group.
>>
>> We show that in many settings this strong notion of protection from
>> discrimination is both attainable and aligned with the goal of obtaining
>> accurate predictions. Along the way, we present new algorithms for learning
>> a multicalibrated predictor, study the computational complexity of this
>> task, and draw new connections to computational learning models such as
>> agnostic learning.
>>
>> Joint work with Ursula Hebert-Johnson, Michael P. Kim and Guy Rothblum
>>
>> Bio:
>> Omer Reingold is a Professor of Computer Science at Stanford University.
>> Past positions include Samsung Research America, the Weizmann Institute of
>> Science, Microsoft Research, the Institute for Advanced Study in Princeton,
>> NJ and AT&T Labs. His research is in the Foundations of Computer Science
>> and most notably in Computational Complexity and the Foundations of
>> Cryptography with emphasis on randomness, derandomization and explicit
>> combinatorial constructions. He has a keen interest in the societal impact
>> of computation. He is an ACM Fellow and among his distinctions are the 2005
>> Grace Murray Hopper Award and the 2009 G?del Prize.
>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180327/34297a32/attachment.html>

From weiyu at cs.cmu.edu  Sat Mar 31 12:19:50 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 31 Mar 2018 09:19:50 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Han Zhao -- April 03
Message-ID: <CABzq7eokz_iWaiYE7=1XOO2tp7EoA2SZ5BZAOiNp-MZ0c9kjow@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, April 03, at noon in *NSH 1507* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday,  Han Zhao <https://www.cs.cmu.edu/~hzhao1/> will give the
following talk:

Title:  Multiple Source Domain Adaptation with Adversarial Learning

Abstract:

While domain adaptation has been actively researched, most theoretical
results and algorithms focus on the single-source-single-target adaptation
setting. In the first part of the talk, I will discuss new generalization
bounds for domain adaptation when there are multiple source domains with
labeled instances and one target domain with unlabeled instances. The
theory also leads to an efficient learning strategy using adversarial
neural networks: I will show how to interpret it as learning feature
representations that are invariant to the multiple domain shifts while
still being discriminative for the learning task.

In the second part, I will discuss two models for multisource domain
adaptations: the first model optimizes the worst-case bound, while the
second model is a smoothed approximation of the first one and optimizes a
task-adaptive bound. We also demonstrate the effectiveness of both models
by conducting extensive experiments showing superior adaptation performance
on three real-world datasets: sentiment analysis, digit classification, and
vehicle counting.

This talk includes joint work with Shanghang Zhang, Guanhang Wu, Joao
Costeira, Jose Moura and Geoff Gordon.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180331/de7ef702/attachment.html>

From weiyu at cs.cmu.edu  Mon Apr  2 06:05:54 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 2 Apr 2018 03:05:54 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Han Zhao -- April
	03
In-Reply-To: <CABzq7eokz_iWaiYE7=1XOO2tp7EoA2SZ5BZAOiNp-MZ0c9kjow@mail.gmail.com>
References: <CABzq7eokz_iWaiYE7=1XOO2tp7EoA2SZ5BZAOiNp-MZ0c9kjow@mail.gmail.com>
Message-ID: <CABzq7eon780uXcdKySTH7F3JCLJNq62m_18f6UwcyhKUh5eYMQ@mail.gmail.com>

A gentle reminder that the talk will happen tomorrow (Tuesday) noon at NSH
1507.


On Sat, Mar 31, 2018 at 9:19 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, April 03, at noon in *NSH
> 1507* for AI Seminar sponsored by Apple. To learn more about the seminar
> series, please visit the AI Seminar webpage
> <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday,  Han Zhao <https://www.cs.cmu.edu/~hzhao1/> will give the
> following talk:
>
> Title:  Multiple Source Domain Adaptation with Adversarial Learning
>
> Abstract:
>
> While domain adaptation has been actively researched, most theoretical
> results and algorithms focus on the single-source-single-target adaptation
> setting. In the first part of the talk, I will discuss new generalization
> bounds for domain adaptation when there are multiple source domains with
> labeled instances and one target domain with unlabeled instances. The
> theory also leads to an efficient learning strategy using adversarial
> neural networks: I will show how to interpret it as learning feature
> representations that are invariant to the multiple domain shifts while
> still being discriminative for the learning task.
>
> In the second part, I will discuss two models for multisource domain
> adaptations: the first model optimizes the worst-case bound, while the
> second model is a smoothed approximation of the first one and optimizes a
> task-adaptive bound. We also demonstrate the effectiveness of both models
> by conducting extensive experiments showing superior adaptation performance
> on three real-world datasets: sentiment analysis, digit classification, and
> vehicle counting.
>
> This talk includes joint work with Shanghang Zhang, Guanhang Wu, Joao
> Costeira, Jose Moura and Geoff Gordon.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180402/2b784bd5/attachment.html>

From weiyu at cs.cmu.edu  Sat Apr  7 17:02:23 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 7 Apr 2018 11:02:23 -1000
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Wen Sun -- April 10
Message-ID: <CABzq7erhoEdxakyL3S7T+KAfEnedjxDRnbTCZwFQ22LxcC9FFw@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, April 10, at noon in NSH 3305 for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday,  Wen Sun <http://www.cs.cmu.edu/~wensun/> will give the
following talk:

Title:  Efficient Reinforcement Learning via Imitation

Abstract:

A fundamental challenge in Artificial Intelligence (AI), robotics, and
language processing is sequential prediction: to reason, plan, and make a
sequence of predictions or decisions to minimize accumulated cost, achieve
a long-term goal, or optimize for a loss acquired only after many
predictions. Reinforcement Learning (RL), as a general framework for
learning from experience to make predictions and decisions, is often
considered as one of the perfect tools for solving such a challenge in AI.
Recently, equipped with the advancement from Deep Learning literature, we
have advanced the state-of-the-art of RL on a number of applications
including simulated high-dimensional robotics control, video games, and
board games (e.g., AlphaGo).

Because of its generality?RL is a general framework that summarizes many
special machine learning algorithms and applications?RL is hard. As there
is no direct supervision, one central challenge in RL is how to explore an
unknown environment and collect useful feedback efficiently. In recent RL
success stories (e.g., super-human performance on video games [Mnih et al.,
2015]), we notice that most of them rely on random exploration strategies,
which usually requires huge number of interactions with the environment
before it can learn anything useful. Another challenge is credit
assignment: if a learning agent successfully achieves some task after
making a long sequence of decisions, how can we assign credit for the
success among these decisions?

We first attempt to gain purchase on RL problems by introducing an
additional source of information?an expert who knows how to solve tasks
(near) optimally. By imitating an expert, we can significantly reduce the
burden of exploration (i.e., we imitate instead of randomly explore), and
solve the credit assignment problem (i.e., the expert tells us which
decisions are bad). We study in both theory and in practice how one can
imitate experts to reduce sample complexity compared to a pure RL approach.

As Imitation Learning is efficient, we next provide a general reduction
from RL to Imitation Learning with a focus on applications where experts
are not available. We explore the possibilities of learning local models
and then using off-shelf model-based RL solvers to compute an intermediate
?expert? for efficient policy improvement via imitation. Furthermore, we
show a general convergence analysis that generalizes and provides the
theoretical foundation for recent successful practical RL algorithms such
as ExIt and AlphaGo Zero [Anthony et al., 2017, Silver et al., 2017], and
provides a theoretical sound and practically efficient way of unifying
model-based and model-free RL approaches.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180407/5d6b1843/attachment.html>

From weiyu at cs.cmu.edu  Mon Apr  9 03:26:01 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 9 Apr 2018 00:26:01 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Wen Sun -- April
	10
In-Reply-To: <CABzq7erhoEdxakyL3S7T+KAfEnedjxDRnbTCZwFQ22LxcC9FFw@mail.gmail.com>
References: <CABzq7erhoEdxakyL3S7T+KAfEnedjxDRnbTCZwFQ22LxcC9FFw@mail.gmail.com>
Message-ID: <CABzq7eq7C3hyvWbsbe5_kwerA-SuQgoACo+JbwKkf29XLWECQw@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in NSH 3305.

On Sat, Apr 7, 2018 at 2:02 PM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, April 10, at noon in NSH 3305 for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday,  Wen Sun <http://www.cs.cmu.edu/~wensun/> will give the
> following talk:
>
> Title:  Efficient Reinforcement Learning via Imitation
>
> Abstract:
>
> A fundamental challenge in Artificial Intelligence (AI), robotics, and
> language processing is sequential prediction: to reason, plan, and make a
> sequence of predictions or decisions to minimize accumulated cost, achieve
> a long-term goal, or optimize for a loss acquired only after many
> predictions. Reinforcement Learning (RL), as a general framework for
> learning from experience to make predictions and decisions, is often
> considered as one of the perfect tools for solving such a challenge in AI.
> Recently, equipped with the advancement from Deep Learning literature, we
> have advanced the state-of-the-art of RL on a number of applications
> including simulated high-dimensional robotics control, video games, and
> board games (e.g., AlphaGo).
>
> Because of its generality?RL is a general framework that summarizes many
> special machine learning algorithms and applications?RL is hard. As there
> is no direct supervision, one central challenge in RL is how to explore an
> unknown environment and collect useful feedback efficiently. In recent RL
> success stories (e.g., super-human performance on video games [Mnih et al.,
> 2015]), we notice that most of them rely on random exploration strategies,
> which usually requires huge number of interactions with the environment
> before it can learn anything useful. Another challenge is credit
> assignment: if a learning agent successfully achieves some task after
> making a long sequence of decisions, how can we assign credit for the
> success among these decisions?
>
> We first attempt to gain purchase on RL problems by introducing an
> additional source of information?an expert who knows how to solve tasks
> (near) optimally. By imitating an expert, we can significantly reduce the
> burden of exploration (i.e., we imitate instead of randomly explore), and
> solve the credit assignment problem (i.e., the expert tells us which
> decisions are bad). We study in both theory and in practice how one can
> imitate experts to reduce sample complexity compared to a pure RL approach.
>
> As Imitation Learning is efficient, we next provide a general reduction
> from RL to Imitation Learning with a focus on applications where experts
> are not available. We explore the possibilities of learning local models
> and then using off-shelf model-based RL solvers to compute an intermediate
> ?expert? for efficient policy improvement via imitation. Furthermore, we
> show a general convergence analysis that generalizes and provides the
> theoretical foundation for recent successful practical RL algorithms such
> as ExIt and AlphaGo Zero [Anthony et al., 2017, Silver et al., 2017], and
> provides a theoretical sound and practically efficient way of unifying
> model-based and model-free RL approaches.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180409/be3667e8/attachment.html>

From weiyu at cs.cmu.edu  Sat Apr 14 06:17:25 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 14 Apr 2018 03:17:25 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Yichong Xu -- April 17
Message-ID: <CABzq7erbXiaaxsCWMH26v8v+zWKo5eedOXWYTjYVAyFUWk48pw@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, April 17, at noon in *NSH 1507* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday,  Yichong Xu <http://xycking.wixsite.com/yichongxu> will give
the following talk:

Title:  Interactive learning using Comparison Queries

Abstract:

In supervised learning, we leverage a labeled dataset to design methods for
function estimation. In many practical situations, we are able to obtain
alternative feedback, possibly at a low cost. A broad goal is to understand
the usefulness of, and to design algorithms to exploit, this alternative
feedback. We consider a interactive learning setting where we obtain
additional ordinal (or comparison) information for potentially unlabeled
samples. In this talk we show the usefulness of such ordinal feedback for
two tasks: Binary classification and nonparametric regression. For binary
classification, we show that comparison queries can help in improving the
label and total query complexity by reducing the learning problem to that
of learning a threshold function. We present an algorithm that achieves
near-optimal label and total query complexity. For nonparametric
regression, we show that it is possible to accurately estimate an
underlying function with a very small labeled set, effectively escaping the
curse of dimensionality. We develop an algorithm called
Ranking-Regression(R^2) and analyze its accuracy as a function of size of
the labeled and unlabeled datasets and various noise parameters. We also
derive lower bounds to show that R^2 is optimal in a variety of settings.
Experiments show that our algorithms outperforms label-only algorithms when
comparison information is available.

Based on joint works with Sivaraman Balakrishnan, Artur Dubrawski, Kyle
Miller, Hariank Muthakana, Aarti Singh and Hongyang Zhang.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180414/5e90c91d/attachment.html>

From weiyu at cs.cmu.edu  Mon Apr 16 07:27:07 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 16 Apr 2018 04:27:07 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Yichong Xu --
	April 17
In-Reply-To: <CABzq7erbXiaaxsCWMH26v8v+zWKo5eedOXWYTjYVAyFUWk48pw@mail.gmail.com>
References: <CABzq7erbXiaaxsCWMH26v8v+zWKo5eedOXWYTjYVAyFUWk48pw@mail.gmail.com>
Message-ID: <CABzq7ep=ZTv_KAGSPiNdfRMeFwYr2YjwmQXh1wxZ7sxnxr+t_g@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in *NSH
1507.*

On Sat, Apr 14, 2018 at 3:17 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, April 17, at noon in *NSH
> 1507* for AI Seminar sponsored by Apple. To learn more about the seminar
> series, please visit the AI Seminar webpage
> <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday,  Yichong Xu <http://xycking.wixsite.com/yichongxu> will give
> the following talk:
>
> Title:  Interactive learning using Comparison Queries
>
> Abstract:
>
> In supervised learning, we leverage a labeled dataset to design methods
> for function estimation. In many practical situations, we are able to
> obtain alternative feedback, possibly at a low cost. A broad goal is to
> understand the usefulness of, and to design algorithms to exploit, this
> alternative feedback. We consider a interactive learning setting where we
> obtain additional ordinal (or comparison) information for potentially
> unlabeled samples. In this talk we show the usefulness of such ordinal
> feedback for two tasks: Binary classification and nonparametric regression.
> For binary classification, we show that comparison queries can help in
> improving the label and total query complexity by reducing the learning
> problem to that of learning a threshold function. We present an algorithm
> that achieves near-optimal label and total query complexity. For
> nonparametric regression, we show that it is possible to accurately
> estimate an underlying function with a very small labeled set, effectively
> escaping the curse of dimensionality. We develop an algorithm called
> Ranking-Regression(R^2) and analyze its accuracy as a function of size of
> the labeled and unlabeled datasets and various noise parameters. We also
> derive lower bounds to show that R^2 is optimal in a variety of settings.
> Experiments show that our algorithms outperforms label-only algorithms when
> comparison information is available.
>
> Based on joint works with Sivaraman Balakrishnan, Artur Dubrawski, Kyle
> Miller, Hariank Muthakana, Aarti Singh and Hongyang Zhang.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180416/f2333b79/attachment.html>

From anupamg at cs.cmu.edu  Wed Apr 18 11:55:17 2018
From: anupamg at cs.cmu.edu (Anupam Gupta)
Date: Wed, 18 Apr 2018 11:55:17 -0400
Subject: [AI Seminar] Theory Lunch: Jonah Sherman on Breaking the l_infinity
	Regularization Barrier
Message-ID: <CAJoAWSU+qcPAYrT2z40=GBV505R+JrpMHGhOx5iyJNdMqv_VEg@mail.gmail.com>

Hi all: talk of potential interest by Jonah Sherman, starting in 5 minutes.

---------- Forwarded message ----------
From: Roie Levin <roiel at andrew.cmu.edu>
Date: Wed, Apr 18, 2018 at 10:15 AM
Subject: Today at Theory Lunch: Jonah Sherman
To: theory-announce at cs.cmu.edu

Hello all,

Please join us Today at noon in GHC 6115 where lunch will be provided. A
video recording of the talk will be available on the

*CMU Youtube Theory channel
<https://www.youtube.com/channel/UCWFp4UWNiOv71j0sPbdNiqw>. *
*Location/Time:* GHC 6115/Today, April 18, 12-1pm

*Speaker:** Jonah Sherman*
*Title**:* Breaking the l_infinity Regularization Barrier: Approximating
Undirected Multicommodity Flow in Nearly-Linear Time

*Abstract: *
Regularization is one of the most powerful tools in continuous
optimization, yet existing approaches using strong-convexity fail for
several important problems due to the infamous "l_infinity barrier". In
this talk, we show strong-convexity may be relaxed to a weaker notion of
"area-convexity", for which those barriers do not apply. Using area-convex
regularization, we obtain a fast algorithm for approximately solving matrix
inequality systems AX <= B over right-stochastic matrices X. By combining
that algorithm with recent work on maximum-flow, we obtain a nearly-linear
time approximation algorithm for maximum concurrent flow in undirected
graphs.

See you there!
Ellis + Roie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180418/d09df1bb/attachment.html>

From weiyu at cs.cmu.edu  Mon Apr 23 01:01:37 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sun, 22 Apr 2018 22:01:37 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Hieu Pham -- April 24
Message-ID: <CABzq7epz4Cgo3D7_sC=e+c=uKFPD-xidHmHm9jMpSWby33YW3Q@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, April 24, at noon in NSH 3305 for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Hieu Pham will give the following talk:

Title: From Neural Combinatorial Optimization to Automatic Machine Learning

Abstract:

Despite neural networks' impressive performance on many tasks, designing
efficient and deploying neural networks stands an arduous challenge. It
involves making a lot of discrete decisions, such as which models to use
and how to deploy such models. Automatizing these designs thus comes at a
great benefit. In this talk, I will show how one can view the task of
designing and deploying neural networks as a combinatorial optimization
problem. Then, I will discuss an application of deep reinforcement learning
(DRL) on a canonical combinatorial optimization task: the Traveling
Salesman Problem (TSP), which outperforms many existing heuristics.
Finally, I will connect the dots, showing that the same DRL approach on TSP
can be applied to automatize the process of designing and deploying neural
networks. Time permits, I will discuss the existing challenges and some
future directions in this line of work.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180422/f14f5fc0/attachment.html>

From weiyu at cs.cmu.edu  Mon Apr 23 22:46:40 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 23 Apr 2018 19:46:40 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Hieu Pham --
	April 24
In-Reply-To: <CABzq7epz4Cgo3D7_sC=e+c=uKFPD-xidHmHm9jMpSWby33YW3Q@mail.gmail.com>
References: <CABzq7epz4Cgo3D7_sC=e+c=uKFPD-xidHmHm9jMpSWby33YW3Q@mail.gmail.com>
Message-ID: <CABzq7epigX6ZLvQRfjdyP5y1edGjPtMOadYZMKxg_j7YzccQaQ@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in NSH 3305.

On Sun, Apr 22, 2018 at 10:01 PM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, April 24, at noon in NSH 3305 for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Hieu Pham will give the following talk:
>
> Title: From Neural Combinatorial Optimization to Automatic Machine
> Learning
>
> Abstract:
>
> Despite neural networks' impressive performance on many tasks, designing
> efficient and deploying neural networks stands an arduous challenge. It
> involves making a lot of discrete decisions, such as which models to use
> and how to deploy such models. Automatizing these designs thus comes at a
> great benefit. In this talk, I will show how one can view the task of
> designing and deploying neural networks as a combinatorial optimization
> problem. Then, I will discuss an application of deep reinforcement learning
> (DRL) on a canonical combinatorial optimization task: the Traveling
> Salesman Problem (TSP), which outperforms many existing heuristics.
> Finally, I will connect the dots, showing that the same DRL approach on TSP
> can be applied to automatize the process of designing and deploying neural
> networks. Time permits, I will discuss the existing challenges and some
> future directions in this line of work.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180423/891d5dff/attachment.html>

From weiyu at cs.cmu.edu  Sat Apr 28 04:19:07 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sat, 28 Apr 2018 01:19:07 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Hongyang Zhang -- May
	01
Message-ID: <CABzq7eouNZpanYK+eA+GxX12OHRFQbqhVFCkAGYqFxSjA3j9nA@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, May 01, at noon in NSH 3305 for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Hongyang Zhang <http://www.cs.cmu.edu/~hongyanz/> will give the
following talk:

Title: Testing and Learning from Big Data, Optimally

Abstract: We are now in an era of big data as well as high-dimensional
data: data volume is almost doubling every two years. Fortunately,
high-dimensional data are structured. Usually, they are of low rank. This
is the basis of dimensionality reduction and compressed sensing. To extract
efficient information from the low-rank structure without fully observing
the matrix, there are two questions to handle: 1. what is the true rank of
the data matrix with only small samples (testing problem)? 2. given the
rank of the matrix, how to design computationally efficient algorithm to
recover the matrix with only compressed observations (learning problem)?

In this talk, we will focus on the testing and learning problems regarding
the matrix rank with optimal sample complexity, which are new paradigms of
information extraction from the big data. In the first part of the talk, we
will see how we can test the rank of an unknown matrix via an interesting
ladder-shaped sampling scheme. We also supplement our positive results with
a hardness result, showing that our sampling scheme is near-optimal.

In the second part of the talk, we study the matrix completion problem.
Matrix completion is known as a non-convex problem in its most original
form. To alleviate the computational issue, we show that strong duality
holds for the matrix completion with nearly optimal sample complexity. For
the hardness result, we also show that generic matrix factorization
requires exponential time to be solved.

Based on joint work with Nina Balcan (CMU), Yi Li (Nanyang Technological
University), Yingyu Liang (Wisconsin-Madison), and David P. Woodruff (CMU).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180428/eb27a406/attachment.html>

From weiyu at cs.cmu.edu  Mon Apr 30 07:42:31 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 30 Apr 2018 04:42:31 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Hongyang Zhang --
	May 01
In-Reply-To: <CABzq7eouNZpanYK+eA+GxX12OHRFQbqhVFCkAGYqFxSjA3j9nA@mail.gmail.com>
References: <CABzq7eouNZpanYK+eA+GxX12OHRFQbqhVFCkAGYqFxSjA3j9nA@mail.gmail.com>
Message-ID: <CABzq7eqhGV4Uu6JxqSBNFs9K=b6RT_P5Bo+wZ0iyifwX=Uunkg@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in NSH 3305.

On Sat, Apr 28, 2018 at 1:19 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, May 01, at noon in NSH 3305 for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Hongyang Zhang <http://www.cs.cmu.edu/~hongyanz/> will give
> the following talk:
>
> Title: Testing and Learning from Big Data, Optimally
>
> Abstract: We are now in an era of big data as well as high-dimensional
> data: data volume is almost doubling every two years. Fortunately,
> high-dimensional data are structured. Usually, they are of low rank. This
> is the basis of dimensionality reduction and compressed sensing. To extract
> efficient information from the low-rank structure without fully observing
> the matrix, there are two questions to handle: 1. what is the true rank of
> the data matrix with only small samples (testing problem)? 2. given the
> rank of the matrix, how to design computationally efficient algorithm to
> recover the matrix with only compressed observations (learning problem)?
>
> In this talk, we will focus on the testing and learning problems regarding
> the matrix rank with optimal sample complexity, which are new paradigms of
> information extraction from the big data. In the first part of the talk, we
> will see how we can test the rank of an unknown matrix via an interesting
> ladder-shaped sampling scheme. We also supplement our positive results with
> a hardness result, showing that our sampling scheme is near-optimal.
>
> In the second part of the talk, we study the matrix completion problem.
> Matrix completion is known as a non-convex problem in its most original
> form. To alleviate the computational issue, we show that strong duality
> holds for the matrix completion with nearly optimal sample complexity. For
> the hardness result, we also show that generic matrix factorization
> requires exponential time to be solved.
>
> Based on joint work with Nina Balcan (CMU), Yi Li (Nanyang Technological
> University), Yingyu Liang (Wisconsin-Madison), and David P. Woodruff (CMU).
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180430/8dd2751b/attachment.html>

From weiyu at cs.cmu.edu  Sun May  6 14:51:33 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sun, 6 May 2018 11:51:33 -0700
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Adams Wei Yu -- May 08
Message-ID: <CABzq7epofAV8nr2vJDo-vqYBq6cHQAh4k-oeeu81kR98b6y+Sg@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, May 08, at noon in *GHC 6115* for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Adams Wei Yu <http://www.cs.cmu.edu/~weiyu/> will give the
following talk:

Title: Efficient and Effective Models for Machine Reading Comprehension

Abstract:

Machine reading comprehension has attracted lots of attentions in the AI,
ML and NLP communities. In this talk, I will introduce two efficient and
effective models to approach this task.

Firstly, I will propose a model, LSTM-Jump, that can skip unimportant
information in sequential data, mimicking the skimming behavior of human
reading. Trained with an efficient reinforcement learning algorithm, this
model can be several times faster than a vanilla LSTM in inference time.

Then I will introduce a sequence encoding method that discards recurrent
networks, which thus fully supports parallel training and inference. Based
on this technique, a new question-answering model, QANet, is proposed.
Combined with data augmentation approach via backtranslation, this model
achieves No.1 performance in the competitive Stanford Question and Answer
Dataset (SQuAD), while being times faster than the prevalent models.
Notably, the exact match score of QANet has exceeded human performance.

The talk is based on the following two works:
1. http://aclweb.org/anthology/P17-1172
2. https://arxiv.org/pdf/1804.09541.pdf
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180506/1546d1d9/attachment.html>

From weiyu at cs.cmu.edu  Mon May  7 16:44:18 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 7 May 2018 16:44:18 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Adams Wei Yu --
	May 08
In-Reply-To: <CABzq7epofAV8nr2vJDo-vqYBq6cHQAh4k-oeeu81kR98b6y+Sg@mail.gmail.com>
References: <CABzq7epofAV8nr2vJDo-vqYBq6cHQAh4k-oeeu81kR98b6y+Sg@mail.gmail.com>
Message-ID: <CABzq7ep9Q5Rw7aTg-R=jJWjmxxDHZ1XGgF5TjysC7fa8uia3Lg@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in *GHC
6115.*

On Sun, May 6, 2018 at 2:51 PM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, May 08, at noon in *GHC 6115* for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Adams Wei Yu <http://www.cs.cmu.edu/~weiyu/> will give the
> following talk:
>
> Title: Efficient and Effective Models for Machine Reading Comprehension
>
> Abstract:
>
> Machine reading comprehension has attracted lots of attentions in the AI,
> ML and NLP communities. In this talk, I will introduce two efficient and
> effective models to approach this task.
>
> Firstly, I will propose a model, LSTM-Jump, that can skip unimportant
> information in sequential data, mimicking the skimming behavior of human
> reading. Trained with an efficient reinforcement learning algorithm, this
> model can be several times faster than a vanilla LSTM in inference time.
>
> Then I will introduce a sequence encoding method that discards recurrent
> networks, which thus fully supports parallel training and inference. Based
> on this technique, a new question-answering model, QANet, is proposed.
> Combined with data augmentation approach via backtranslation, this model
> achieves No.1 performance in the competitive Stanford Question and Answer
> Dataset (SQuAD), while being times faster than the prevalent models.
> Notably, the exact match score of QANet has exceeded human performance.
>
> The talk is based on the following two works:
> 1. http://aclweb.org/anthology/P17-1172
> 2. https://arxiv.org/pdf/1804.09541.pdf
>
>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180507/b0528bef/attachment.html>

From weiyu at cs.cmu.edu  Sun May 13 05:09:00 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Sun, 13 May 2018 05:09:00 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Guoqing Zheng -- May
	15
Message-ID: <CABzq7erXyunYufe-zNcXnuZ8VowBK46Ca30fBP4_LQdWmyVvEg@mail.gmail.com>

Dear faculty and students,

We look forward to seeing you next Tuesday, May 15, at noon in NSH 3305 for
AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.

On Tuesday, Guoqing Zheng <http://www.cs.cmu.edu/~gzheng/> will give the
following talk:

Title: Generative Adversarial Permutation Learning

Abstract:

Permutation learning refers to the task of recovering the mapping that
permutes an object to another one. The (original object, permuted object)
pair encodes information that is critical in identifying the underlying
permutation and has received increasingly attention in the research
community. In this talk, we study the problem of unpaired permutation
learning, i.e., only samples of original objects and that of permuted
objects are observed while no paired link between the two is given. We
propose to tackle the unpaired permutation learning under the adversarial
training framework; specifically, a permutation generative network is
trained to generate approximated permutations conditioned on the permuted
object, by which the permuted object can be transformed back to the
original space, and a discriminative network is trained to distinguish real
objects from the original space and the recovered ones. Preliminary
empirical experiments on sorting numbers and recovering scrambled images
demonstrates the effectiveness of the proposed method.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180513/2f6911e4/attachment.html>

From weiyu at cs.cmu.edu  Mon May 14 07:40:47 2018
From: weiyu at cs.cmu.edu (Adams Wei Yu)
Date: Mon, 14 May 2018 07:40:47 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Guoqing Zheng --
	May 15
In-Reply-To: <CABzq7erXyunYufe-zNcXnuZ8VowBK46Ca30fBP4_LQdWmyVvEg@mail.gmail.com>
References: <CABzq7erXyunYufe-zNcXnuZ8VowBK46Ca30fBP4_LQdWmyVvEg@mail.gmail.com>
Message-ID: <CABzq7epzH0PpMAMEGZt3f1+88ERdV=dRTNG4pQDL3aH3H5LeGw@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon in NSH 3305.

On Sun, May 13, 2018 at 5:09 AM, Adams Wei Yu <weiyu at cs.cmu.edu> wrote:

> Dear faculty and students,
>
> We look forward to seeing you next Tuesday, May 15, at noon in NSH 3305 for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the AI Seminar webpage <http://www.cs.cmu.edu/~aiseminar/>.
>
> On Tuesday, Guoqing Zheng <http://www.cs.cmu.edu/~gzheng/> will give the
> following talk:
>
> Title: Generative Adversarial Permutation Learning
>
> Abstract:
>
> Permutation learning refers to the task of recovering the mapping that
> permutes an object to another one. The (original object, permuted object)
> pair encodes information that is critical in identifying the underlying
> permutation and has received increasingly attention in the research
> community. In this talk, we study the problem of unpaired permutation
> learning, i.e., only samples of original objects and that of permuted
> objects are observed while no paired link between the two is given. We
> propose to tackle the unpaired permutation learning under the adversarial
> training framework; specifically, a permutation generative network is
> trained to generate approximated permutations conditioned on the permuted
> object, by which the permuted object can be transformed back to the
> original space, and a discriminative network is trained to distinguish real
> objects from the original space and the recovered ones. Preliminary
> empirical experiments on sorting numbers and recovering scrambled images
> demonstrates the effectiveness of the proposed method.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180514/dd4a1a44/attachment.html>

From han.zhao at cs.cmu.edu  Wed Aug 29 10:41:44 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Wed, 29 Aug 2018 10:41:44 -0400
Subject: [AI Seminar] [AI-Seminar] Call for Talks in AI Seminar, Fall 2018
Message-ID: <CANHBZyDEd1yMrj6Mre8i=c0YwgMyq7f8PwxWuiQnq+jTXvSieA@mail.gmail.com>

Dear all:

Welcome back to school and hope you all had a great summer!

The CMU AI Seminar <http://www.cs.cmu.edu/~aiseminar/> this semester will
start from Tuesday, Sep. 11th. Please sign up for contributing a talk!
Again, based on previous experience, the slots will be filled up soon, so
please do it as soon as possible you can! The tentative schedule for this
semester is listed in the website <http://www.cs.cmu.edu/~aiseminar/> .

Looking forward to seeing you at the seminars soon!
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180829/60a8d69f/attachment.html>

From han.zhao at cs.cmu.edu  Fri Sep  7 14:56:38 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Fri, 7 Sep 2018 14:56:38 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Matt Barnes -- Sep.
	11th
Message-ID: <CANHBZyCyqS8j00HLGogo_Wo_B-j4dpys1PgTOKeOqTZ+491P4w@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Sep. 11th, at noon in
*GHC 6115* for AI Seminar sponsored by Apple. To learn more about the
seminar series, please visit the website <http://www.cs.cmu.edu/~aiseminar/>
.
On Tuesday, Matt Barnes <http://www.cs.cmu.edu/~mbarnes1/> will give the
following talk:

Title: Learning with Clusters: A cardinal machine learning sin and how to
correct for it

Abstract: As machine learning systems become increasingly complex,
clustering has evolved from an exploratory data analysis tool into an
integrated component of computer vision, robotics, medical and census data
pipelines. Currently, as with many machine learning systems, the output of
the clustering algorithm is taken as ground truth at the next pipeline
step. We show this false assumption causes subtle and dangerous behavior
for even the simplest systems -- sometimes biasing results by upwards of
25%.

We provide the first empirical and theoretical study of this phenomenon
which we term dependency leakage. Further, we introduce fixes in the form
of estimators and methods to both quantify and correct for clustering
errors' impacts on downstream learners. Our work is agnostic to the
downstream learners, and requires few assumptions on the clustering
algorithm. Empirical results demonstrate our approach improves these
machine learning systems compared to naive approaches, which do not account
for clustering errors.

This talk is based on the following two papers:
http://auai.org/uai2017/proceedings/papers/87.pdf
https://arxiv.org/abs/1807.06713

-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180907/db6aaa8c/attachment.html>

From han.zhao at cs.cmu.edu  Mon Sep 10 11:12:32 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 10 Sep 2018 11:12:32 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Matt Barnes --
	Sep. 11th
In-Reply-To: <CANHBZyCyqS8j00HLGogo_Wo_B-j4dpys1PgTOKeOqTZ+491P4w@mail.gmail.com>
References: <CANHBZyCyqS8j00HLGogo_Wo_B-j4dpys1PgTOKeOqTZ+491P4w@mail.gmail.com>
Message-ID: <CANHBZyB0dOrS17uE1jiLfkJ3WWhhp54Q6o8DVDfpJ7w-oh=pYA@mail.gmail.com>

A gentle reminder that the talk will be tomorrow (Tuesday) noon at *GHC
6115*. See you then!

Han Zhao <han.zhao at cs.cmu.edu> ?2018?9?7??? ??2:56???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Sep. 11th, at noon in
> *GHC 6115* for AI Seminar sponsored by Apple. To learn more about the
> seminar series, please visit the website
> <http://www.cs.cmu.edu/~aiseminar/>.
> On Tuesday, Matt Barnes <http://www.cs.cmu.edu/~mbarnes1/> will give the
> following talk:
>
> Title: Learning with Clusters: A cardinal machine learning sin and how to
> correct for it
>
> Abstract: As machine learning systems become increasingly complex,
> clustering has evolved from an exploratory data analysis tool into an
> integrated component of computer vision, robotics, medical and census data
> pipelines. Currently, as with many machine learning systems, the output of
> the clustering algorithm is taken as ground truth at the next pipeline
> step. We show this false assumption causes subtle and dangerous behavior
> for even the simplest systems -- sometimes biasing results by upwards of
> 25%.
>
> We provide the first empirical and theoretical study of this phenomenon
> which we term dependency leakage. Further, we introduce fixes in the form
> of estimators and methods to both quantify and correct for clustering
> errors' impacts on downstream learners. Our work is agnostic to the
> downstream learners, and requires few assumptions on the clustering
> algorithm. Empirical results demonstrate our approach improves these
> machine learning systems compared to naive approaches, which do not account
> for clustering errors.
>
> This talk is based on the following two papers:
> http://auai.org/uai2017/proceedings/papers/87.pdf
> https://arxiv.org/abs/1807.06713
>
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180910/fdbc4bcb/attachment.html>

From han.zhao at cs.cmu.edu  Sun Sep 16 00:56:12 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sun, 16 Sep 2018 00:56:12 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Benjamin Eysenbach --
	Sep. 18th
Message-ID: <CANHBZyDe9aXTEZ3=V6hDooP2hdFGjvmE-Y-WNs+_JHJGAb+eJA@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Sep. 18th, at noon in
*NSH 3305* for AI Seminar sponsored by Apple. To learn more about the
seminar series, please visit the website <http://www.cs.cmu.edu/~aiseminar/>
.
On Tuesday, Benjamin Eysenbach <https://ben-eysenbach.github.io/> will give
the following talk:

Title: Towards Autonomous Reinforcement Learning: Learning to Act with Less
Human Supervision

Abstract: Widespread adoption of RL today is severely limited by its
dependence on human supervision. In this talk, I'll discuss many areas
where current approaches to RL requires human supervision. A couple recent
projects make progress on this problem by partially removing the this
dependence, enabling more autonomous RL. I'll conclude with some thoughts
on how RL could be made even more autonomous.

This talk is largely based off the following papers:
Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement
Learning <https://arxiv.org/abs/1711.06782>
Diversity is All You Need: Learning Skills without a Reward Function
<https://arxiv.org/abs/1802.06070>
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180916/c8168f7a/attachment.html>

From han.zhao at cs.cmu.edu  Mon Sep 17 09:21:03 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 17 Sep 2018 09:21:03 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Benjamin Eysenbach --
	Sep. 18th
Message-ID: <CANHBZyBa4EJJGk-q2=K924Af5FNg2By3pqEdpFjgoDLSUs=N0w@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you on Tuesday, Sep. 18th, at noon in
*NSH 3305* for AI Seminar sponsored by Apple. To learn more about the
seminar series, please visit the website <http://www.cs.cmu.edu/~aiseminar/>
.
On Tuesday, Benjamin Eysenbach <https://ben-eysenbach.github.io/> will give
the following talk:

Title: Towards Autonomous Reinforcement Learning: Learning to Act with Less
Human Supervision

Abstract: Widespread adoption of RL today is severely limited by its
dependence on human supervision. In this talk, I'll discuss many areas
where current approaches to RL requires human supervision. A couple recent
projects make progress on this problem by partially removing the this
dependence, enabling more autonomous RL. I'll conclude with some thoughts
on how RL could be made even more autonomous.

This talk is largely based off the following papers:
Leave no Trace: Learning to Reset for Safe and Autonomous Reinforcement
Learning <https://arxiv.org/abs/1711.06782>
Diversity is All You Need: Learning Skills without a Reward Function
<https://arxiv.org/abs/1802.06070>
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180917/3dfe96e9/attachment.html>

From han.zhao at cs.cmu.edu  Tue Sep 18 11:12:40 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Tue, 18 Sep 2018 11:12:40 -0400
Subject: [AI Seminar] AI Seminar Series: Jakub Pachocki (OpenAI), Sep 25th,
	2018
Message-ID: <CANHBZyDFAwX_MDi270-7yV=mj=gm-dKOA2_wTUojn+SEvi9H_Q@mail.gmail.com>

Dear faculty and students:

Jakub Pachocki (OpenAI) will be presenting at the AI Seminar Series, Sep
25, 2018 12:00 PM GHC 6115.
http://www.cs.cmu.edu/~aiseminar/

Please schedule meetings with Dr. Pachocki using the schedule below if you
are interested:
https://docs.google.com/spreadsheets/d/1fXQ4ZUkQ2eNeJF111Kz4jvvlF_vrnhY29cJircgR2gM/edit?usp=sharing
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180918/3c77d52a/attachment.html>

From han.zhao at cs.cmu.edu  Sat Sep 22 13:34:51 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sat, 22 Sep 2018 13:34:51 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Jakub Pachocki --
	Sep. 25th
Message-ID: <CANHBZyA0MutYN+2Peem=RpSEN77V39sFukyYnrYX6uOu4wzbCQ@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Sep. 25th, at noon in
*GHC 6115* for AI Seminar sponsored by Apple. To learn more about the
seminar series, please visit the website <http://www.cs.cmu.edu/~aiseminar/>
.
On Tuesday, Jakub Pachocki <https://www.linkedin.com/in/jakub-pachocki/> will
give the following talk:

Title: OpenAI Five: Training a Dota-2 Bot through Self-play

Abstract: Dota 2 is a popular, competitive, real-time strategy game played
in two teams of five players. I will discuss our path, through scaling
reinforcement learning to thousands of GPUs and hundreds of thousands of
CPUS, to creating a deep learning based AI competitive with professionals.
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180922/9551d8e3/attachment.html>

From han.zhao at cs.cmu.edu  Mon Sep 24 12:51:12 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 24 Sep 2018 12:51:12 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Jakub Pachocki --
	Sep. 25th
In-Reply-To: <CANHBZyA0MutYN+2Peem=RpSEN77V39sFukyYnrYX6uOu4wzbCQ@mail.gmail.com>
References: <CANHBZyA0MutYN+2Peem=RpSEN77V39sFukyYnrYX6uOu4wzbCQ@mail.gmail.com>
Message-ID: <CANHBZyAmPtH1HAQ6qFQ7s4XQeLLNp2Gv8PL3M=STtgqYuAr5DA@mail.gmail.com>

A gentle reminder that the following talk will open tomorrow (Tuesday) noon
at *GHC 6115*.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?9?22??? ??1:34???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Sep. 25th, at noon in
> *GHC 6115* for AI Seminar sponsored by Apple. To learn more about the
> seminar series, please visit the website
> <http://www.cs.cmu.edu/~aiseminar/>.
> On Tuesday, Jakub Pachocki <https://www.linkedin.com/in/jakub-pachocki/> will
> give the following talk:
>
> Title: OpenAI Five: Training a Dota-2 Bot through Self-play
>
> Abstract: Dota 2 is a popular, competitive, real-time strategy game
> played in two teams of five players. I will discuss our path, through
> scaling reinforcement learning to thousands of GPUs and hundreds of
> thousands of CPUS, to creating a deep learning based AI competitive with
> professionals.
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180924/51d4cd8d/attachment.html>

From han.zhao at cs.cmu.edu  Sat Sep 29 17:58:01 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sat, 29 Sep 2018 17:58:01 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Gus Guangyu Xia --
	Oct. 2nd
Message-ID: <CANHBZyAjk8u42pmNiuzr642mPnJFfZhAVYgdcCLAVsmgw8BDGw@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Oct. 2nd, at noon in GHC 6115
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Gus Guangyu Xia will give the following talk:

Title: Towards More Creative and Interactive Music AI

Abstract: Gus works on music AI systems for intimate human-computer
interaction and co-creation. In this talk, he will present two recent
prototype systems: the haptic-flute tutor and the popsong piano arranger.
The former is a hyperinstrument which can guild human fingers and make
music learning more efficient and enjoyable. The latter is a general
framwork which unifies melody and accompaniment creation. At the end of the
talk, Gus will show his vision on merging these two efforts for new music
experiences.

-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20180929/63f52f91/attachment.html>

From han.zhao at cs.cmu.edu  Mon Oct  1 12:34:02 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 1 Oct 2018 12:34:02 -0400
Subject: [AI Seminar] Fwd: AI Seminar sponsored by Apple -- Gus Guangyu Xia
	-- Oct. 2nd
In-Reply-To: <CANHBZyAjk8u42pmNiuzr642mPnJFfZhAVYgdcCLAVsmgw8BDGw@mail.gmail.com>
References: <CANHBZyAjk8u42pmNiuzr642mPnJFfZhAVYgdcCLAVsmgw8BDGw@mail.gmail.com>
Message-ID: <CANHBZyBtXbg3DkTie86CZOn1vhXV97S6Lzs3dtHF38y20iB4Vg@mail.gmail.com>

A gentle reminder that the following talk will open tomorrow (Tuesday) noon
at *GHC 6115*.

---------- Forwarded message ---------
From: Han Zhao <han.zhao at cs.cmu.edu>
Date: 2018?9?29??? ??5:58
Subject: AI Seminar sponsored by Apple -- Gus Guangyu Xia -- Oct. 2nd
To: <ai-seminar-announce at cs.cmu.edu>


Dear faculty and students:

We look forward to seeing you next Tuesday, Oct. 2nd, at noon in GHC 6115
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Gus Guangyu Xia will give the following talk:

Title: Towards More Creative and Interactive Music AI

Abstract: Gus works on music AI systems for intimate human-computer
interaction and co-creation. In this talk, he will present two recent
prototype systems: the haptic-flute tutor and the popsong piano arranger.
The former is a hyperinstrument which can guild human fingers and make
music learning more efficient and enjoyable. The latter is a general
framwork which unifies melody and accompaniment creation. At the end of the
talk, Gus will show his vision on merging these two efforts for new music
experiences.

-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181001/62653ee3/attachment.html>

From han.zhao at cs.cmu.edu  Sat Oct  6 16:19:53 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sat, 6 Oct 2018 16:19:53 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Qizhe Xie -- Oct. 9th
Message-ID: <CANHBZyCgPO6vZkVH3faevFkoptw6xzid1V_83C9E=LzRrqK89Q@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Oct. 9th, at noon in GHC 6115
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Qizhe Xie will give the following talk:

Title: From Credit Assignment to Entropy Regularization: Two New Algorithms
for Neural Sequence Prediction

Background:
Modeling and predicting discrete sequences is the central problem to many
natural language processing tasks. Despite the distinct evaluation metrics
for different tasks, the standard training algorithm for language
generation has been maximum likelihood estimation (MLE). However, the MLE
algorithm has two obvious weaknesses: (1) the MLE training ignores the
information of the task specific metric; (2) MLE can suffer from the
exposure bias, which refers to the phenomenon that the model is never
exposed to its own failures during training. The recently proposed reward
augmented maximum likelihood (RAML) tackles these problems by constructing
a task metric dependent target distribution, and training the model to
match this task-specific target instead of the empirical data distribution.

Abstract:
In this talk, we study the credit assignment problem in reward augmented
maximum likelihood (RAML), and establish a theoretical equivalence between
the token-level counterpart of RAML and the entropy regularized
reinforcement learning. Inspired by the connection, we propose two sequence
prediction algorithms, one extending RAML with fine-grained credit
assignment and the other improving Actor-Critic with a systematic entropy
regularization. On two benchmark datasets, we show that the proposed
algorithms outperform RAML and Actor-Critic respectively.

-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181006/06ee352f/attachment.html>

From han.zhao at cs.cmu.edu  Mon Oct  8 15:19:03 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 8 Oct 2018 15:19:03 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Qizhe Xie -- Oct.
	9th
In-Reply-To: <CANHBZyCgPO6vZkVH3faevFkoptw6xzid1V_83C9E=LzRrqK89Q@mail.gmail.com>
References: <CANHBZyCgPO6vZkVH3faevFkoptw6xzid1V_83C9E=LzRrqK89Q@mail.gmail.com>
Message-ID: <CANHBZyCQimSH98FAZ5DfHmpj6Eg+uNodDqnxunRe+vyaCOvjNA@mail.gmail.com>

A gentle reminder that the following talk will happen at noon in GHC 6115
tomorrow.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?10?6??? ??4:19???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Oct. 9th, at noon in GHC 6115
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the website.
> On Tuesday, Qizhe Xie will give the following talk:
>
> Title: From Credit Assignment to Entropy Regularization: Two New
> Algorithms for Neural Sequence Prediction
>
> Background:
> Modeling and predicting discrete sequences is the central problem to many
> natural language processing tasks. Despite the distinct evaluation metrics
> for different tasks, the standard training algorithm for language
> generation has been maximum likelihood estimation (MLE). However, the MLE
> algorithm has two obvious weaknesses: (1) the MLE training ignores the
> information of the task specific metric; (2) MLE can suffer from the
> exposure bias, which refers to the phenomenon that the model is never
> exposed to its own failures during training. The recently proposed reward
> augmented maximum likelihood (RAML) tackles these problems by constructing
> a task metric dependent target distribution, and training the model to
> match this task-specific target instead of the empirical data distribution.
>
> Abstract:
> In this talk, we study the credit assignment problem in reward augmented
> maximum likelihood (RAML), and establish a theoretical equivalence between
> the token-level counterpart of RAML and the entropy regularized
> reinforcement learning. Inspired by the connection, we propose two sequence
> prediction algorithms, one extending RAML with fine-grained credit
> assignment and the other improving Actor-Critic with a systematic entropy
> regularization. On two benchmark datasets, we show that the proposed
> algorithms outperform RAML and Actor-Critic respectively.
>
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181008/b2520524/attachment.html>

From han.zhao at cs.cmu.edu  Sun Oct 14 12:38:04 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sun, 14 Oct 2018 12:38:04 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Sarah Keren -- Oct.
	16th
Message-ID: <CANHBZyCrvadmZUDp=TkCr2oA10197hW5aTqkRVSTtKUDycTKgA@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Oct. 16th, at noon in NSH 3305
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Sarah Keren will give the following talk:

Title: Goal Recognition Design

Abstract: Goal recognition design (GRD) is the task of redesigning
environments in order to facilitate online goal recognition. As such, while
goal recognition tools are typically aimed at efficiently analyzing online
observations of agents (human or automated) in order to infer their
objective, GRD focuses on manipulating the environment in which agents act
to guarantee early recognition.

In a nutshell, given a model of a domain and a set of possible goals, a
solution to a GRD problem determines: (1) to what extent do actions,
performed by an agent within the model, reveal the agent???s objective? and
(2) what is the best way to modify the model so that the objective of an
agent is revealed as early as possible? GRD answers these questions by
offering a solution for assessing and minimizing the maximal progress of an
agent in the model before its goal is revealed. This approach is relevant
to any domain for which quickly performing goal recognition is essential
and in which the model design can be controlled. Applications include
intrusion detection, assisted cognition, computer games, and human-robot
collaboration.

Using several motivating examples, my talk will cover the models and
methods created to asses and optimize various GRD settings. In addition, I
will present recent work which extends the redesign approach to settings
with arbitrary utility measures. The utility maximizing design (UMD)
framework brings new exciting directions to explore, such as formulating
the design process as a heuristic search and finding informative heuristics
to guide the search for optimal design strategies.

Bio: Sarah Keren is a postdoctoral fellow at Harvard University, where she
is affiliated with the Center for Research on Computation and Society
(CRCS). Her mentors are Prof. Barbara Grosz and Prof. David Parkes. Before
coming to Harvard, Sarah completed her Ph.D. at the Faculty of Industrial
Engineering and Management of the Technion - Israel Institute of
Technology, where she was advised by Prof. Avigdor Gal and Dr. Erez Karpas.
Sarah's research focuses on manipulating and redesigning environments for
optimizing their utility. Her work has appeared in three leading artificial
intelligence conferences (AAAI, ICAPS and IJCAI). She has received
different excellence awards including an honorable mention for best paper
in ICAPS 2014, as well as the Eric and Wendy Schmidt Postdoctoral Award for
Women in Mathematical and Computing Sciences.
--
Han Zhao
Machine Learning Department
School of Computer Science
Carnegie Mellon University
Mobile: +1-412-652-4404
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181014/9acce3d6/attachment.html>

From han.zhao at cs.cmu.edu  Tue Oct 16 11:11:52 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Tue, 16 Oct 2018 11:11:52 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Sarah Keren --
	Oct. 16th
In-Reply-To: <CANHBZyCrvadmZUDp=TkCr2oA10197hW5aTqkRVSTtKUDycTKgA@mail.gmail.com>
References: <CANHBZyCrvadmZUDp=TkCr2oA10197hW5aTqkRVSTtKUDycTKgA@mail.gmail.com>
Message-ID: <CANHBZyCXs94+F+UiFn4M0dx8aWC0Awbcm+L6eqpD5b_ETs-BNw@mail.gmail.com>

A gentile reminder that the following talk will happen today 12 pm at NSH
3305.


Han Zhao <han.zhao at cs.cmu.edu> ?2018?10?14??? ??12:38???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Oct. 16th, at noon in NSH 3305
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the website.
> On Tuesday, Sarah Keren will give the following talk:
>
> Title: Goal Recognition Design
>
> Abstract: Goal recognition design (GRD) is the task of redesigning
> environments in order to facilitate online goal recognition. As such, while
> goal recognition tools are typically aimed at efficiently analyzing online
> observations of agents (human or automated) in order to infer their
> objective, GRD focuses on manipulating the environment in which agents act
> to guarantee early recognition.
>
> In a nutshell, given a model of a domain and a set of possible goals, a
> solution to a GRD problem determines: (1) to what extent do actions,
> performed by an agent within the model, reveal the agent???s objective? and
> (2) what is the best way to modify the model so that the objective of an
> agent is revealed as early as possible? GRD answers these questions by
> offering a solution for assessing and minimizing the maximal progress of an
> agent in the model before its goal is revealed. This approach is relevant
> to any domain for which quickly performing goal recognition is essential
> and in which the model design can be controlled. Applications include
> intrusion detection, assisted cognition, computer games, and human-robot
> collaboration.
>
> Using several motivating examples, my talk will cover the models and
> methods created to asses and optimize various GRD settings. In addition, I
> will present recent work which extends the redesign approach to settings
> with arbitrary utility measures. The utility maximizing design (UMD)
> framework brings new exciting directions to explore, such as formulating
> the design process as a heuristic search and finding informative heuristics
> to guide the search for optimal design strategies.
>
> Bio: Sarah Keren is a postdoctoral fellow at Harvard University, where she
> is affiliated with the Center for Research on Computation and Society
> (CRCS). Her mentors are Prof. Barbara Grosz and Prof. David Parkes. Before
> coming to Harvard, Sarah completed her Ph.D. at the Faculty of Industrial
> Engineering and Management of the Technion - Israel Institute of
> Technology, where she was advised by Prof. Avigdor Gal and Dr. Erez Karpas.
> Sarah's research focuses on manipulating and redesigning environments for
> optimizing their utility. Her work has appeared in three leading artificial
> intelligence conferences (AAAI, ICAPS and IJCAI). She has received
> different excellence awards including an honorable mention for best paper
> in ICAPS 2014, as well as the Eric and Wendy Schmidt Postdoctoral Award for
> Women in Mathematical and Computing Sciences.
> --
> Han Zhao
> Machine Learning Department
> School of Computer Science
> Carnegie Mellon University
> Mobile: +1-412-652-4404
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181016/ef515801/attachment.html>

From han.zhao at cs.cmu.edu  Sat Oct 20 21:50:15 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sat, 20 Oct 2018 21:50:15 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Devendra Chaplot --
	Oct. 23rd
Message-ID: <CANHBZyDrUjcSf_AD+HatU0H448hMVCo6HTpCWH=dBu3+T6fDTA@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Oct. 23rd, at noon in GHC 6115
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Devendra Chaplot will give the following talk:

Title: Embodied Multimodal Multitask Learning

Abstract: Recent efforts on training visual navigation agents conditioned
on language using deep reinforcement learning have been successful in
learning policies for two multimodal tasks: learning to follow navigational
instructions and embodied question answering. We aim to learn a multitask
model capable of jointly learning both tasks, and transferring knowledge of
words and their grounding in visual objects across tasks. The proposed
model uses a novel Dual-Attention unit to disentangle the knowledge of
words in the textual representations and visual objects in the visual
representations, and align them with each other. This disentangled
task-invariant alignment of representations facilitates grounding and
knowledge transfer across both tasks. We show that the proposed model
outperforms a range of baselines on both tasks in simulated 3D
environments. We also show that this disentanglement of representations
makes our model modular, interpretable, and allows for zero-shot transfer
to instructions containing new words by leveraging object detectors.
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181020/ffc1bdb5/attachment.html>

From han.zhao at cs.cmu.edu  Mon Oct 22 14:39:49 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 22 Oct 2018 14:39:49 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Devendra Chaplot
	-- Oct. 23rd
In-Reply-To: <CANHBZyDrUjcSf_AD+HatU0H448hMVCo6HTpCWH=dBu3+T6fDTA@mail.gmail.com>
References: <CANHBZyDrUjcSf_AD+HatU0H448hMVCo6HTpCWH=dBu3+T6fDTA@mail.gmail.com>
Message-ID: <CANHBZyD5Tb1Un=H7QAGk+8xoGPTHnqfE54VEPzEYu9_yTwkQVw@mail.gmail.com>

A gentile reminder that the following talk will happen tomorrow 12 pm at
GHC 6115.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?10?20??? ??9:50???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Oct. 23rd, at noon in GHC 6115
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the website.
> On Tuesday, Devendra Chaplot will give the following talk:
>
> Title: Embodied Multimodal Multitask Learning
>
> Abstract: Recent efforts on training visual navigation agents conditioned
> on language using deep reinforcement learning have been successful in
> learning policies for two multimodal tasks: learning to follow navigational
> instructions and embodied question answering. We aim to learn a multitask
> model capable of jointly learning both tasks, and transferring knowledge of
> words and their grounding in visual objects across tasks. The proposed
> model uses a novel Dual-Attention unit to disentangle the knowledge of
> words in the textual representations and visual objects in the visual
> representations, and align them with each other. This disentangled
> task-invariant alignment of representations facilitates grounding and
> knowledge transfer across both tasks. We show that the proposed model
> outperforms a range of baselines on both tasks in simulated 3D
> environments. We also show that this disentanglement of representations
> makes our model modular, interpretable, and allows for zero-shot transfer
> to instructions containing new words by leveraging object detectors.
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181022/f1467a9f/attachment.html>

From han.zhao at cs.cmu.edu  Sat Oct 27 23:25:20 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sat, 27 Oct 2018 23:25:20 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Renato Negrinho --
	Oct. 30th
Message-ID: <CANHBZyDbMEZR2CbG+SUACMu4Yb3tsXE8EpPZ4YCSt6+khnVPdg@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Oct. 30th, at noon in NSH 3305
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Renato Negrinho will give the following talk:

*Title: *Learning Beam Search Policies via Imitation Learning
*Abstract: *Beam search is widely used for approximate decoding in
structured prediction problems. Models often use a beam at test time but
ignore its existence at train time, and therefore do not explicitly learn
how to use the beam. % The Solution: our meta-algorithm We develop an
unifying meta-algorithm for learning beam search policies using imitation
learning. In our setting, the beam is part of the model, and not just an
artifact of approximate decoding. Our meta-algorithm captures existing
learning algorithms and suggests new ones. It also lets us show novel
no-regret guarantees for learning beam search policies.
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181027/68da7989/attachment.html>

From han.zhao at cs.cmu.edu  Tue Oct 30 09:39:45 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Tue, 30 Oct 2018 09:39:45 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Renato Negrinho
	-- Oct. 30th
In-Reply-To: <CANHBZyDbMEZR2CbG+SUACMu4Yb3tsXE8EpPZ4YCSt6+khnVPdg@mail.gmail.com>
References: <CANHBZyDbMEZR2CbG+SUACMu4Yb3tsXE8EpPZ4YCSt6+khnVPdg@mail.gmail.com>
Message-ID: <CANHBZyCvXKvo8frgVD3L+sBW-RBsaGzujFhwuegcax9UMA8gZA@mail.gmail.com>

A gentle reminder that the following talk will happen 12pm today at NSH
3305.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?10?27??? ??11:25???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Oct. 30th, at noon in NSH 3305
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the website.
> On Tuesday, Renato Negrinho will give the following talk:
>
> *Title: *Learning Beam Search Policies via Imitation Learning
> *Abstract: *Beam search is widely used for approximate decoding in
> structured prediction problems. Models often use a beam at test time but
> ignore its existence at train time, and therefore do not explicitly learn
> how to use the beam. % The Solution: our meta-algorithm We develop an
> unifying meta-algorithm for learning beam search policies using imitation
> learning. In our setting, the beam is part of the model, and not just an
> artifact of approximate decoding. Our meta-algorithm captures existing
> learning algorithms and suggests new ones. It also lets us show novel
> no-regret guarantees for learning beam search policies.
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181030/460a1151/attachment.html>

From han.zhao at cs.cmu.edu  Sun Nov  4 00:05:52 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sun, 4 Nov 2018 00:05:52 -0400
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Yifan Wu -- Nov. 6th
Message-ID: <CANHBZyBsyaixQK0jJKLSfYiaex7iBE7PYwx0Eh5Uw=uyzm+BEA@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Nov. 6th, at noon in GHC 6115
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Yifan Wu will give the following talk:

*Title: *The Laplacian in Reinforcement Learning: Learning Representations
without Rewards and Decoders

*Abstract: *The smallest eigenvectors of the graph Laplacian are well-known
to provide a succinct representation of the geometry of a weighted graph.
In reinforcement learning (RL), where the weighted graph may be interpreted
as the state transition process induced by a behavior policy acting on the
environment, approximating the eigenvectors of the Laplacian provides a
promising approach to state representation learning. However, existing
methods for performing this approximation are ill-suited in general RL
settings for two main reasons:  First, they are computationally expensive,
often requiring operations on large matrices. Second, these methods lack
adequate justification beyond simple, tabular, finite-state settings. In
this paper, we present a fully general and scalable method for
approximating the eigenvectors of the Laplacian in a model-free RL context.
We systematically evaluate our approach and empirically show that it
generalizes beyond the tabular, finite-state setting. Even in tabular,
finite-state settings, its ability to approximate the eigenvectors
outperforms previous proposals. Finally, we show the potential benefits of
using a Laplacian representation learned using our method in goal-achieving
RL tasks, providing evidence that our technique can be used to
significantly improve the performance of an RL agent.
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181104/fed6699a/attachment.html>

From han.zhao at cs.cmu.edu  Sun Nov 11 21:05:26 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sun, 11 Nov 2018 21:05:26 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Kirstin Early -- Nov.
	13th
Message-ID: <CANHBZyCMGHmOXs6-b9KwnLbvbKXg4tqwUqtbEy=ArP1dJAtjZg@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Nov. 13th, at noon in GHC 6115
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Kirstin Early will give the following talk:

*Title: Towards a general purpose text representation for mail*

*Abstract: *There are many possibilities for using machine learning on
email to help users save time and accomplish their goals -- e.g., spam
classification, reply prediction, and message categorization. However,
building a separate model for each task is inefficient because each model
generates its own internal representation of messages. A general-purpose
representation could eliminate this redundant computation and be used for
many downstream tasks. We train encoder-decoder neural networks on
self-supervised mail tasks and generate representations of new mail
messages as the encoder output of these networks. Simple models for
downstream tasks can then be trained on the representations. We illustrate
this method on a pilot task of RSVP classification and find the
general-purpose representation performs similarly to a model built
specifically for this task.
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181111/a3b42e57/attachment.html>

From han.zhao at cs.cmu.edu  Tue Nov 13 09:32:40 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Tue, 13 Nov 2018 09:32:40 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Kirstin Early --
	Nov. 13th
In-Reply-To: <CANHBZyCMGHmOXs6-b9KwnLbvbKXg4tqwUqtbEy=ArP1dJAtjZg@mail.gmail.com>
References: <CANHBZyCMGHmOXs6-b9KwnLbvbKXg4tqwUqtbEy=ArP1dJAtjZg@mail.gmail.com>
Message-ID: <CANHBZyCUE9oP7LhhhLt6O5oCHD_s4gHSmknSbWMri+rHDKRkAQ@mail.gmail.com>

A gentle reminder that the following talk will happen today noon at GHC
6115.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?11?11??? ??9:05???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Nov. 13th, at noon in GHC 6115
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the website.
> On Tuesday, Kirstin Early will give the following talk:
>
> *Title: Towards a general purpose text representation for mail*
>
> *Abstract: *There are many possibilities for using machine learning on
> email to help users save time and accomplish their goals -- e.g., spam
> classification, reply prediction, and message categorization. However,
> building a separate model for each task is inefficient because each model
> generates its own internal representation of messages. A general-purpose
> representation could eliminate this redundant computation and be used for
> many downstream tasks. We train encoder-decoder neural networks on
> self-supervised mail tasks and generate representations of new mail
> messages as the encoder output of these networks. Simple models for
> downstream tasks can then be trained on the representations. We illustrate
> this method on a pilot task of RSVP classification and find the
> general-purpose representation performs similarly to a model built
> specifically for this task.
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181113/6f6a4897/attachment.html>

From han.zhao at cs.cmu.edu  Sat Nov 17 12:12:52 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sat, 17 Nov 2018 12:12:52 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Jonathan Laurent --
	Nov. 20th
Message-ID: <CANHBZyCKDOTPOuf6xc5x45+YaVYGNApWrUxsx+m0j5cSjAd3LA@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Nov. 20th, at noon in NSH 3305
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Jonathan Laurent will give the following talk:

*Title: Counterfactual Resimulation for Causal Analysis of Rule-Based
Models*

*Abstract: *A major challenge in Systems Biology is to uncover the causal
mechanisms through which complex cell behaviors emerge from a multitude of
low-level molecular interactions. Understanding these causal mechanisms is
critical for medical research as it enables the identification of potential
targets for drugs.

In this spirit, biologists have been building models of the cell chemistry
that aggregate known interactions between proteins. In this talk, we
introduce a technique to analyze the causal structure of simulation traces
generated from such models. Concretely, our method leverages simulation to
produce causal diagrams that explain outcomes of interest in terms of
enablement and prevention relations between molecular events.

A key contribution of our work is an algorithm to simulate counterfactual
scenarios efficiently. Given a reference trace, it provides an answer to
questions such as: What would have happened had a particular event not
happened? Doing so, it generalizes Pearl's standard treatment of
counterfactuals to complex dynamical systems.

(Based on joint work with Jean Yang and Walter Fontana, published at
IJCAI-18.)
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181117/24ede8b9/attachment.html>

From han.zhao at cs.cmu.edu  Mon Nov 19 10:52:02 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 19 Nov 2018 10:52:02 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Jonathan Laurent
	-- Nov. 20th
In-Reply-To: <CANHBZyCKDOTPOuf6xc5x45+YaVYGNApWrUxsx+m0j5cSjAd3LA@mail.gmail.com>
References: <CANHBZyCKDOTPOuf6xc5x45+YaVYGNApWrUxsx+m0j5cSjAd3LA@mail.gmail.com>
Message-ID: <CANHBZyAekT5Pz5YQ73hKWBs4VP+4L0Lt9VZuCgmKBTBkgkpeUw@mail.gmail.com>

A gentle reminder that the following talk will happen tomorrow noon at NSH
3305.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?11?17??? ??12:12???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Nov. 20th, at noon in NSH 3305
> for AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the website.
> On Tuesday, Jonathan Laurent will give the following talk:
>
> *Title: Counterfactual Resimulation for Causal Analysis of Rule-Based
> Models*
>
> *Abstract: *A major challenge in Systems Biology is to uncover the causal
> mechanisms through which complex cell behaviors emerge from a multitude of
> low-level molecular interactions. Understanding these causal mechanisms is
> critical for medical research as it enables the identification of potential
> targets for drugs.
>
> In this spirit, biologists have been building models of the cell chemistry
> that aggregate known interactions between proteins. In this talk, we
> introduce a technique to analyze the causal structure of simulation traces
> generated from such models. Concretely, our method leverages simulation to
> produce causal diagrams that explain outcomes of interest in terms of
> enablement and prevention relations between molecular events.
>
> A key contribution of our work is an algorithm to simulate counterfactual
> scenarios efficiently. Given a reference trace, it provides an answer to
> questions such as: What would have happened had a particular event not
> happened? Doing so, it generalizes Pearl's standard treatment of
> counterfactuals to complex dynamical systems.
>
> (Based on joint work with Jean Yang and Walter Fontana, published at
> IJCAI-18.)
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181119/016d46d6/attachment.html>

From han.zhao at cs.cmu.edu  Sun Nov 25 15:12:34 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sun, 25 Nov 2018 15:12:34 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Yaodong Yu -- Nov.
	27th
Message-ID: <CANHBZyBFR3uWN7ijzCkoWFhJRTh3brkqcie2-2kVdXL5nPqr=Q@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Nov. 27th, at noon in GHC 6115
for AI Seminar sponsored by Apple. To learn more about the seminar series,
please visit the website.
On Tuesday, Yaodong Yu will give the following talk:

*Title: Adversarial Defenses and Attacks: A Case Study on NIPS 2018
Adversarial Vision Challenge*

*Abstract: *In recent years, we have witnessed the success of machine
learning, especially deep learning, in various areas. Despite widespread
adoption, recent studies have shown that machine learning models are
vulnerable to adversarial examples, i.e., very small changes to inputs can
cause machine learning models to make mistakes. Thus, understanding and
defending against adversarial examples is crucial to the AI security and
interpretability concerns.

In this talk, we will focus on how to train robust models and generate
adversarial examples in the NIPS 2018 Adversarial Vision Challenge. In the
first part of the talk, we will briefly introduce basic adversarial defense
and attack techniques as well as the rule in the NIPS 2018 Adversarial
Vision Challenge. In the second part of the talk, we will present several
take-home messages on how to train robust models efficiently on large-scale
datasets (Tiny ImageNet) by using deep convolutional neural networks (e.g.,
ResNet152). In addition, we will present our approach on generating
adversarial examples for targeted attacks.

Based on joint work with Hongyang Zhang (CMU), Susu Xu (CMU), Hongbao Zhang
(Petuum), Pengtao Xie (Petuum) and Eric P. Xing (CMU and Petuum).
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181125/2c0a58d7/attachment.html>

From han.zhao at cs.cmu.edu  Mon Dec  3 12:42:22 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 3 Dec 2018 12:42:22 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Ivan Stelmakh -- Dec.
	4th
Message-ID: <CANHBZyAmsH_UXzAr7Uy-Zy6aL_a_7tsjiEDtHYCCQdYhFHoSLA@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you tomorrow, Dec. 4th, at noon in GHC 6115 for AI
 Seminar sponsored by Apple. To learn more about the seminar series, please
visit the website.
On Tuesday, Ivan Stelmakh will give the following talk:

*Title: PeerReview4All: Fair and Accurate Reviewer Assignment in Peer
Review*
*Abstract: *In this talk, we will focus on automated assignment of papers
to reviewers in conference peer review. In the first part of the talk we
will show that assignment procedure currently employed by NeurIPS and ICML
does not guarantee fairness. Optimizing the total quality of the assignment
over all papers, this procedure may discriminate against some submissions.
In contrast, we will present the assignment algorithm that maximizes the
review quality of the most disadvantaged paper, thus ensuring fairness. In
the second part of talk, we will discuss the parallel objective of accuracy
and show that under standard statistical model, our algorithm leads to the
minimax optimal accuracy of the final decisions. Finally, we will present a
novel experiment that allows for an objective comparison of various
assignment algorithms in terms of both fairness and accuracy, and overcomes
the inherent difficulty posed by the absence of a ground truth in
experiments on peer-review. The results of this experiment corroborate the
theoretical guarantees of our algorithm.
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181203/ba8510dc/attachment.html>

From han.zhao at cs.cmu.edu  Tue Dec  4 10:28:11 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Tue, 4 Dec 2018 10:28:11 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Ivan Stelmakh --
	Dec. 4th
In-Reply-To: <CANHBZyAmsH_UXzAr7Uy-Zy6aL_a_7tsjiEDtHYCCQdYhFHoSLA@mail.gmail.com>
References: <CANHBZyAmsH_UXzAr7Uy-Zy6aL_a_7tsjiEDtHYCCQdYhFHoSLA@mail.gmail.com>
Message-ID: <CANHBZyA55iMNGwcrMWVnZksxpyD+2Q-6mhx5Axn+GAf6g4Aqrg@mail.gmail.com>

A gentle reminder that the following talk will happen at noon *today* in
GHC 6115.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?12?3??? ??12:42???

> Dear faculty and students:
>
> We look forward to seeing you tomorrow, Dec. 4th, at noon in GHC 6115 for
> AI Seminar sponsored by Apple. To learn more about the seminar series,
> please visit the website.
> On Tuesday, Ivan Stelmakh will give the following talk:
>
> *Title: PeerReview4All: Fair and Accurate Reviewer Assignment in Peer
> Review*
> *Abstract: *In this talk, we will focus on automated assignment of papers
> to reviewers in conference peer review. In the first part of the talk we
> will show that assignment procedure currently employed by NeurIPS and ICML
> does not guarantee fairness. Optimizing the total quality of the assignment
> over all papers, this procedure may discriminate against some submissions.
> In contrast, we will present the assignment algorithm that maximizes the
> review quality of the most disadvantaged paper, thus ensuring fairness. In
> the second part of talk, we will discuss the parallel objective of accuracy
> and show that under standard statistical model, our algorithm leads to the
> minimax optimal accuracy of the final decisions. Finally, we will present a
> novel experiment that allows for an objective comparison of various
> assignment algorithms in terms of both fairness and accuracy, and overcomes
> the inherent difficulty posed by the absence of a ground truth in
> experiments on peer-review. The results of this experiment corroborate the
> theoretical guarantees of our algorithm.
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181204/7cfdec49/attachment.html>

From han.zhao at cs.cmu.edu  Sun Dec  9 11:18:45 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sun, 9 Dec 2018 11:18:45 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Rediet Abebe -- Dec.
	11th
Message-ID: <CANHBZyC2KsQWBN-V7y9QYsVpn+ktqAT1UmPSuq4b19tupzd-yA@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Dec. 11th, at noon in *Univ.
Center, Danforth Conference Room *for AI Seminar sponsored by Apple. To
learn more about the seminar series, please visit the website.
On Tuesday, Rediet Abebe will give the following talk:

*Title: **Computational Interventions to Improve Access to Opportunity for
Disadvantaged Populations*

*Abstract: *Poverty and economic hardship are highly complex and dynamic
phenomena. Due to the multi-dimensional nature of economic welfare,
assistance programs aimed at improving access to opportunity for
disadvantaged populations face challenges when relevant information about
these populations is unavailable or (even when such information is
available) when they are forced to rely on simplistic measures of welfare
(e.g., household income or wealth). In this presentation, we explore
algorithmic and computational challenges that arise in this settings.

In the first part of the talk, we explore one important dimension of
economic welfare: susceptibility to income shocks in the form of an
unexpected bill or disruption of one's income flow. We introduce and
analyze a model of economic welfare that incorporates income, wealth, and
external shocks and poses the question of how to allocate subsidies in this
setting. We find that we can give optimal allocation mechanisms under
natural assumptions on the agent's wealth and shock distributions, as well
as approximation schemes in the general setting.

In the second part of the talk, we consider settings in which relevant
information -- such as individuals' information needs -- is not available.
Focusing on the lack of comprehensive, high-quality data about the health
information needs of individuals in developing nations, we propose a
bottom-up approach that uses search data from individuals in all 54 nations
in Africa. We analyze Bing searches related to HIV/AIDS, malaria, and
tuberculosis; these searches reveal diverse health information needs that
vary by demographic groups and geographic regions. We also shed light on
discrepancies in the quality of content returned by search engines.

We conclude with a discussion on how algorithmic, computational, and
mechanism design techniques can help inform interventions to improve access
to opportunity in relevant domains and the Mechanism Design for Social Good
research initiative.

This talk is based on joint work with Shawndra Hill, Jon Kleinberg, H.
Andrew Schwartz, Peter M. Small, Jennifer Wortman Vaughan, and S. Matthew
Weinberger.

*Bio*: Rediet Abebe is a Ph.D. candidate in computer science at Cornell
University, advised by Professor Jon Kleinberg. Her research focuses on
algorithms, AI, and applications to social good. She uses computational
insights to improve access to opportunity, with a focus on under-served and
marginalized communities. As part of this research mission, she co-founded
and co-organizes the Mechanism Design for Social Good (MD4SG) initiative,
an interdisciplinary, multi-institutional research group. She is also a
co-founder and co-organizer of Black in AI, a transcontinental group aimed
at increasing the presence and inclusion of Black researchers in the field
of AI. Her research is deeply influenced by her upbringing in her hometown
of Addis Ababa, Ethiopia, where she lived until moving to the U.S. in 2009.
Her work has been generously supported by fellowships and scholarships
through Facebook, Google, the Cornell Graduate School, and the
Harvard-Cambridge Fellowship.
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181209/2c37ec95/attachment.html>

From han.zhao at cs.cmu.edu  Mon Dec 10 10:31:54 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Mon, 10 Dec 2018 10:31:54 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Rediet Abebe --
	Dec. 11th
In-Reply-To: <CANHBZyC2KsQWBN-V7y9QYsVpn+ktqAT1UmPSuq4b19tupzd-yA@mail.gmail.com>
References: <CANHBZyC2KsQWBN-V7y9QYsVpn+ktqAT1UmPSuq4b19tupzd-yA@mail.gmail.com>
Message-ID: <CANHBZyA-xDhKxGU-jparacBfp+MTf8pWRsFd9j2-kwxGs+ut-w@mail.gmail.com>

Dear faculty and students:

Rediet will be on campus for the whole day tomorrow, please sign up a time
slot on the google sheet (
https://docs.google.com/spreadsheets/d/1oUZGO2VqE1F2uqsWmeu6HgyPRmT6nPkxyuRdOojDVRU/edit#gid=0)
if you'd like to meet and talk with her.

Best,
Han.

Han Zhao <han.zhao at cs.cmu.edu> ?2018?12?9??? ??11:18???

> Dear faculty and students:
>
> We look forward to seeing you next Tuesday, Dec. 11th, at noon in *Univ.
> Center, Danforth Conference Room *for AI Seminar sponsored by Apple. To
> learn more about the seminar series, please visit the website.
> On Tuesday, Rediet Abebe will give the following talk:
>
> *Title: **Computational Interventions to Improve Access to Opportunity
> for Disadvantaged Populations*
>
> *Abstract: *Poverty and economic hardship are highly complex and dynamic
> phenomena. Due to the multi-dimensional nature of economic welfare,
> assistance programs aimed at improving access to opportunity for
> disadvantaged populations face challenges when relevant information about
> these populations is unavailable or (even when such information is
> available) when they are forced to rely on simplistic measures of welfare
> (e.g., household income or wealth). In this presentation, we explore
> algorithmic and computational challenges that arise in this settings.
>
> In the first part of the talk, we explore one important dimension of
> economic welfare: susceptibility to income shocks in the form of an
> unexpected bill or disruption of one's income flow. We introduce and
> analyze a model of economic welfare that incorporates income, wealth, and
> external shocks and poses the question of how to allocate subsidies in this
> setting. We find that we can give optimal allocation mechanisms under
> natural assumptions on the agent's wealth and shock distributions, as well
> as approximation schemes in the general setting.
>
> In the second part of the talk, we consider settings in which relevant
> information -- such as individuals' information needs -- is not available.
> Focusing on the lack of comprehensive, high-quality data about the health
> information needs of individuals in developing nations, we propose a
> bottom-up approach that uses search data from individuals in all 54 nations
> in Africa. We analyze Bing searches related to HIV/AIDS, malaria, and
> tuberculosis; these searches reveal diverse health information needs that
> vary by demographic groups and geographic regions. We also shed light on
> discrepancies in the quality of content returned by search engines.
>
> We conclude with a discussion on how algorithmic, computational, and
> mechanism design techniques can help inform interventions to improve access
> to opportunity in relevant domains and the Mechanism Design for Social Good
> research initiative.
>
> This talk is based on joint work with Shawndra Hill, Jon Kleinberg, H.
> Andrew Schwartz, Peter M. Small, Jennifer Wortman Vaughan, and S. Matthew
> Weinberger.
>
> *Bio*: Rediet Abebe is a Ph.D. candidate in computer science at Cornell
> University, advised by Professor Jon Kleinberg. Her research focuses on
> algorithms, AI, and applications to social good. She uses computational
> insights to improve access to opportunity, with a focus on under-served and
> marginalized communities. As part of this research mission, she co-founded
> and co-organizes the Mechanism Design for Social Good (MD4SG) initiative,
> an interdisciplinary, multi-institutional research group. She is also a
> co-founder and co-organizer of Black in AI, a transcontinental group aimed
> at increasing the presence and inclusion of Black researchers in the field
> of AI. Her research is deeply influenced by her upbringing in her hometown
> of Addis Ababa, Ethiopia, where she lived until moving to the U.S. in 2009.
> Her work has been generously supported by fellowships and scholarships
> through Facebook, Google, the Cornell Graduate School, and the
> Harvard-Cambridge Fellowship.
> --
>
> *Han ZhaoMachine Learning Department*
>
>
> *School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
> *412-652-4404*
>


-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181210/77e63cec/attachment.html>

From han.zhao at cs.cmu.edu  Sun Dec 16 14:26:18 2018
From: han.zhao at cs.cmu.edu (Han Zhao)
Date: Sun, 16 Dec 2018 14:26:18 -0500
Subject: [AI Seminar] AI Seminar sponsored by Apple -- Po-Wei Wang -- Dec.
	18th
Message-ID: <CANHBZyDjW96n9_Z-oi1ZWMASfUxrzZmm+xu3PspC19i1STQM=Q@mail.gmail.com>

Dear faculty and students:

We look forward to seeing you next Tuesday, Dec. 18th, at noon in *NSH
3305 *for AI Seminar sponsored by Apple. To learn more about the
seminar series,
please visit the website.
On Tuesday, Po-Wei Wang will give the following talk:

*Title: **Low-rank Semidefinite Programming for the MAX2SAT Problem*

*Abstract: *This paper proposes a new algorithm for solving MAX2SAT
problems based on combining search methods with semidefinite programming
approaches. Semidefinite programming techniques are well-known as a
theoretical tool for approximating maximum satisfiability problems, but
their application has traditionally been very limited by their speed and
randomized nature. Our approach overcomes this difficult by using a recent
approach to low-rank semidefinite programming, specialized to work in an
incremental fashion suitable for use in an exact search algorithm. The
method can be used both within complete or incomplete solver, and we
demonstrate on a variety of problems from recent competitions. Our
experiments show that the approach is faster (sometimes by orders of
magnitude) than existing state-of-the-art complete and incomplete solvers,
representing a substantial advance in search methods specialized for
MAX2SAT problems.

(Based on joint work with J. Zico Kolter, published at AAAI-19.)
-- 

*Han ZhaoMachine Learning Department*


*School of Computer ScienceCarnegie Mellon UniversityMobile: +1-*
*412-652-4404*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/ai-seminar-announce/attachments/20181216/ec22ef51/attachment.html>