Two Auton Lab thesis defenses next week! Mark your calendars please...

[Artur Dubrawski]

And the details of Maria's defense on Monday:

Please join us on Monday, May 4 via Zoom at 11am when Maria De-Arteaga (ML
& Public Policy Joint PhD) will be defending her thesis.

*Title:* Machine Learning in High-Stakes Settings: Risks and Opportunities

*Thesis committee:* Artur Dubrawski (Co-Chair), Alexandra Chouldechova
(Co-Chair), Roni Rosenfeld, Adam Tauman Kalai (Microsoft Research)

*Zoom Link:*
https://cmu.zoom.us/j/94967473449?pwd=b09lL29qblg1ZU5BWHZhVDB2NjFjQT09

*Meeting ID:* 949 6747 3449
*Password:* 000312

*Abstract: * Machine learning (ML) is increasingly being used to support
decision-making in critical settings, where predictions have potentially
grave implications over human lives. Examples include healthcare, hiring,
child welfare, and the criminal justice system. In this thesis, I study the
risks and opportunities of machine learning in high-stakes settings. In the
first chapter I focus on opportunities of ML to support experts' decisions
when dealing with high-resolution multivariate data, a type of data that is
particularly hard for humans to interpret. I propose methodology to
discover latent complex multivariate correlation structures and illustrate
its use in two different domains: (1) identification of radioactive threats
in nuclear physics, and (2) prediction of neurological recovery of comatose
patients in healthcare. In the second chapter, focused on algorithmic
fairness, I demonstrate how societal biases encoded in historical data may
be reproduced and amplified by ML models, and introduce a new algorithm to
mitigate biases without assuming access to protected attributes. Finally,
in the third chapter I characterize challenges that arise from the
limitations of available labels in decision support contexts--such as the
selective labels problem and omitted payoff bias--and propose methodology
to estimate and leverage human consistency to improve algorithmic
recommendations and human-machine complementarity.


*Paper Link:*
https://www.dropbox.com/s/h449z85r6nls8oc/Dissertation_DeArteaga.pdf?dl=0


On Tue, Apr 28, 2020 at 11:28 AM Artur Dubrawski <awd at cs.cmu.edu> wrote:

> Dear Autonians,
>
> Please join me in attending 2 (yes, two) excellent virtual
> presentations by our own Maria De-Arteaga and Chao Liu, both of which are
> scheduled for the next week.
>
> (btw, I do not remember when was the last time we had more than one
> doctoral thesis defense scheduled in one week at the Auton Lab...)
>
> Maria's defense will be on Monday May 4th at 11am,
> The official announcement will be shared soon.
>
> Chao's defense is scheduled for Thursday May 7th at noon.
> The official announcement with the zoom link is included below.
>
> Please help seeing these outstanding colleagues move to the next levels of
> their professional lives by attending these presentations and cheering for
> them :)
>
> Cheers,
> Artur
>
> -----
>
> Date:  07 May 2020
>
> Time:  12:00 p.m.
>
> Place: *Virtual Presentation* https://cmu.zoom.us/j/2623852919
>
> Type:  Ph.D. Thesis Defense
>
> Who:  Chao Liu
>
> Title:  Vision with Small Baselines
>
>
> Abstract:
> 3D sensing with portable imaging systems is becoming more and more popular
> in computer vision applications such as autonomous driving, virtual
> reality, robotics manipulation and surveillance, due to the decreasing
> expense and size of RGB cameras. Despite the compactness and portability of
> the small baseline vision systems, it is well-known that the uncertainty in
> range finding using multiple views and the sensor baselines are inversely
> related. On the other hand, besides compactness, the small baseline vision
> system has its unique advantages such as easier correspondence and large
> overlapping regions across views.
>
> The goal of this thesis is to develop computational methods and small
> baseline imaging systems for 3D sensing of complex scenes in real world
> conditions. Our design principle is to physically model the scene
> complexities and specifically infer the uncertainties for the images
> captured with small baseline setups.
>
> With this design principle, we make four contributions. In the first
> contribution, we propose a two-stage near-light photometric stereo method
> using a small (6 cm diameter) LED ring. The imaging system is compact
> compared to traditional photometric stereo systems. In the second
> contribution, we develop an algorithm to simultaneously estimate the
> occlusion pattern and depth for thin structures from a focal image stack,
> which is obtained either by varying the focus/aperture of the lens or
> computed from a one-shot light field image. As the third contribution, we
> propose a  learning-based method to estimate per-pixel depth and its
> uncertainty continuously from a monocular video stream, with small camera
> baselines across adjacent frames. These depth probability volumes are
> accumulated over time as more incoming frames are processed sequentially,
> which effectively reduces depth uncertainty and improves accuracy,
> robustness, and temporal stability. Finally, using a pair of high
> resolution camera and laser projector, we develop a high spatial resolution
> Diffuse Optical Tomography (DOT) system that can detect accurate boundaries
> and relative depth of heterogeneous structures up to a depth of 8mm below a
> highly scattering medium such as whole milk.
>
> We showcase the application of a small baseline vision system for in-vivo
> micro-scale 3D reconstruction of capillary veins and develop a system for
> real-time analysis of microvascular blood flow for critical care. We
> believe that the computational methods developed in this thesis would find
> more applications of compact 3D sensing under challenging conditions.
>
>
>
> Thesis Committee Members:
>
> Srinivasa G. Narasimhan, Co-chair
> Artur W. Dubrawski, Co-chair
> Aswin C. Sankaranarayanan
> Manmohan Chandraker, University of California, San Diego
>
>
> A copy of the thesis document is available at:
>
> https://www.dropbox.com/s/cz75koh96ragy4x/thesis-small-baseline.pdf?dl=0
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20200429/d1412081/attachment-0001.html>