From predragp at andrew.cmu.edu  Wed Nov 10 12:37:24 2021
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Wed, 10 Nov 2021 12:37:24 -0500
Subject: Unable to login through bash.autonlab.org
In-Reply-To: <CAAT765Xj9PvdpoaX2kH4QUciARwGwOa2Qop5m8OeDCx4HhGzvA@mail.gmail.com>
References: <CAAT765Xj9PvdpoaX2kH4QUciARwGwOa2Qop5m8OeDCx4HhGzvA@mail.gmail.com>
Message-ID: <CAEFFt-GnV-4YEYCzfAeadrd46GtxiaAFRsWeLxE3_X5Lx+Mpzw@mail.gmail.com>

Hi Adam,

Thank you for this report. It appears that the OpenVPN client died not just
on bash but on all the Auton Lab desktops due to the CMU network hiccup.

Please try now both bash and lion. They should work. I am still
investigating what else went wrong.

Predrag


On Wed, Nov 10, 2021 at 10:39 AM Adam Harley <aharley at cmu.edu> wrote:

> Hi,
> Just to let you know: when I try ssh bash.autonlab.org, it asks me for my
> password (instead of recognizing my ssh key), and also my password is not
> working. I was able to login through lop2 though.
> Best,
> Adam
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20211110/84c40756/attachment.html>

From aharley at cmu.edu  Wed Nov 10 12:41:34 2021
From: aharley at cmu.edu (Adam Harley)
Date: Wed, 10 Nov 2021 12:41:34 -0500
Subject: Unable to login through bash.autonlab.org
In-Reply-To: <CAEFFt-GnV-4YEYCzfAeadrd46GtxiaAFRsWeLxE3_X5Lx+Mpzw@mail.gmail.com>
References: <CAAT765Xj9PvdpoaX2kH4QUciARwGwOa2Qop5m8OeDCx4HhGzvA@mail.gmail.com>
 <CAEFFt-GnV-4YEYCzfAeadrd46GtxiaAFRsWeLxE3_X5Lx+Mpzw@mail.gmail.com>
Message-ID: <CAAT765Vy1CJSHde+8RRJnFmPBEUMOQ5pJqdukWXOhV3MDPwxWg@mail.gmail.com>

Everything works for me now. Thanks!

On Wed, Nov 10, 2021 at 12:37 PM Predrag Punosevac <predragp at andrew.cmu.edu>
wrote:

> Hi Adam,
>
> Thank you for this report. It appears that the OpenVPN client died not
> just on bash but on all the Auton Lab desktops due to the CMU network
> hiccup.
>
> Please try now both bash and lion. They should work. I am still
> investigating what else went wrong.
>
> Predrag
>
>
> On Wed, Nov 10, 2021 at 10:39 AM Adam Harley <aharley at cmu.edu> wrote:
>
>> Hi,
>> Just to let you know: when I try ssh bash.autonlab.org, it asks me for
>> my password (instead of recognizing my ssh key), and also my password is
>> not working. I was able to login through lop2 though.
>> Best,
>> Adam
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20211110/a13489ec/attachment.html>

From redman at andrew.cmu.edu  Wed Nov 10 14:43:04 2021
From: redman at andrew.cmu.edu (Robert Edman)
Date: Wed, 10 Nov 2021 14:43:04 -0500
Subject: Research Thrust meetings
Message-ID: <CAM+K4XyAUbw3c4w1cz2UyF0YACaD+odDSy_zis=RO=yVPeBSXA@mail.gmail.com>

Hi All,

Kyle and I are repurposing the Thursday 1pm cardio meeting Kyle created to
start a series of research topic meetings to facilitate collaboration and
technical progress. These are meant to complement the brainstorming session
by focusing on problems we are facing on our current projects, identifying
research needs, and matching those needs with current research. We'll start
with predictive maintenance since our next round of PMx funding is expected
any day now.

Our schedule for the next month is:
11/11 - Predictive Maintenance
11/18 - Nuclear Detection/Physics based modeling
12/2 - Computer Vision
12/10 - Cardiorespiratory health

Everyone is welcome to attend and we'll create some slack channels to
coordinate and announce things.

Expect to hear more in the coming weeks as we figure out how to make these
useful. If you have any questions, comments, or concerns, please reach out
to Kyle or I.

--Rob
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20211110/f4543896/attachment.html>

From awd at cs.cmu.edu  Thu Nov 18 15:26:05 2021
From: awd at cs.cmu.edu (Artur Dubrawski)
Date: Thu, 18 Nov 2021 15:26:05 -0500
Subject: Our in-hospital biosurveillance tool making noise in the media
Message-ID: <CAJvAoysn6_Yjj+M=g_SQM+oORTEdktggY_kEFVM0iYcrRoKzKw@mail.gmail.com>

Dear Autonians,

Our Enhanced Detection System for Healthcare-Associated Transmission
algorithm is getting popular after some really good results came out
from a retrospective impact study led by our collaborators at Pitt.

See for instance this piece from Pittsburgh Post Gazette:
https://www.post-gazette.com/news/health/2021/11/17/University-of-PIttsburgh-CMU-hospital-infection-outbreaks-artificial-intelligence/stories/202111160095
but we should probably expect more noise.

This is another good example of how AI research done at the Auton Lab
can make a tangible impact in the world.

Way to go Auton EDS-HAT Team! (it prominently includes Kyle and Jessie
all the time from the start of the project some six years ago, and
most recently adds Dan, Andrew and Stef)

Cheers,
Artur

From predragp at andrew.cmu.edu  Fri Nov 19 15:22:52 2021
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Fri, 19 Nov 2021 15:22:52 -0500
Subject: GPU1 fixed
Message-ID: <CAEFFt-H+DmWf+uwOw4WcAPDNjHHvOTJfvxbFFqDKgBkfev8tug@mail.gmail.com>

It is fixed and 100% usable. Data is intact. 10 TB RAID6 is clean. FYI the
warranty on this one expired in January of 2019 so we already got +3 years
on the top of standard 5 years. Silicon Mechanics guys were able to get me
somebody from Taiwan to talk to me inspite of warranty status. The issue
was resolved 20 minutes later.

If you care about the issues (PMx guys) please ping me. It is TL:TW

Predrag
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20211119/d6dae11a/attachment.html>

From jeff4 at andrew.cmu.edu  Mon Nov 22 18:22:27 2021
From: jeff4 at andrew.cmu.edu (Jeff Schneider)
Date: Mon, 22 Nov 2021 18:22:27 -0500
Subject: GPU1 fixed
In-Reply-To: <CAEFFt-H+DmWf+uwOw4WcAPDNjHHvOTJfvxbFFqDKgBkfev8tug@mail.gmail.com>
References: <CAEFFt-H+DmWf+uwOw4WcAPDNjHHvOTJfvxbFFqDKgBkfev8tug@mail.gmail.com>
Message-ID: <d6c218f2-86b1-984d-298a-66bb01e07451@andrew.cmu.edu>

Thanks Predrag!!? Amazing to hear you were able to save both machine and 
data!

On 11/19/2021 3:22 PM, Predrag Punosevac wrote:
> It is fixed and 100% usable. Data is intact. 10 TB RAID6 is clean. FYI 
> the warranty on this one expired?in January of 2019 so we already 
> got?+3 years on the top of standard 5 years. Silicon Mechanics guys 
> were able to get me somebody from?Taiwan to talk to me inspite of 
> warranty status. The issue was resolved?20 minutes later.
>
> If you care about the issues (PMx guys) please ping me. It is TL:TW
>
> Predrag

From predragp at andrew.cmu.edu  Tue Nov 23 22:28:07 2021
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Tue, 23 Nov 2021 22:28:07 -0500
Subject: lov4 is inaccessible
In-Reply-To: <CABG36ZZ7WRRmEcgk6URVh6e4emTHzYinG_JxrTmPm3Fp=LxtbQ@mail.gmail.com>
References: <CABG36ZZ7WRRmEcgk6URVh6e4emTHzYinG_JxrTmPm3Fp=LxtbQ@mail.gmail.com>
Message-ID: <CAEFFt-GfwXWsV8Psg8Z=omoCHs=5+XEBQYdLRJsvv6=YePbUuQ@mail.gmail.com>

Hi Arundhati,

I just checked on lov4. It appears that the hardware is completely dead. I
could not reach it from my home even with IPMI (a separate ARM
motherboard). This will have to wait until I get back from the Thanksgiving
break and have physical access to the hardware. I took some time off so I
will be back to CMU next Tuesday.

FYI lov4 is a 9 year old computing node and I would not be surprised if it
turns out that it is not fixable. I can replace a dead power supply, RAM
module or HDD but that's about it. If anything else is wrong with the
server it will be decommissioned.

Cheers,
Predrag


On Tue, Nov 23, 2021 at 10:04 PM Arundhati Banerjee <arundhat at andrew.cmu.edu>
wrote:

> Hi Predrag,
>
> I am currently unable to ssh into lov4 and from monit it seems that lov4
> is down. It would be really helpful if you could please take a look.
>
> Thank you!
> Best regards,
> Arundhati
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20211123/71cf0a1b/attachment.html>

From awd at cs.cmu.edu  Wed Nov 24 09:57:43 2021
From: awd at cs.cmu.edu (Artur Dubrawski)
Date: Wed, 24 Nov 2021 09:57:43 -0500
Subject: EDS-HAT in Wall Street Journal [Our in-hospital biosurveillance tool
 making noise in the media]
In-Reply-To: <CAJvAoysn6_Yjj+M=g_SQM+oORTEdktggY_kEFVM0iYcrRoKzKw@mail.gmail.com>
References: <CAJvAoysn6_Yjj+M=g_SQM+oORTEdktggY_kEFVM0iYcrRoKzKw@mail.gmail.com>
Message-ID: <CAJvAoyt7SeEysuQGwN-RffKW53hbWhn=emZis7_sTiMjykRuuQ@mail.gmail.com>

On Thu, Nov 18, 2021 at 3:26 PM Artur Dubrawski <awd at cs.cmu.edu> wrote:
>
> Dear Autonians,
>
> Our Enhanced Detection System for Healthcare-Associated Transmission
> algorithm is getting popular after some really good results came out
> from a retrospective impact study led by our collaborators at Pitt.
>
> See for instance this piece from Pittsburgh Post Gazette:
> https://www.post-gazette.com/news/health/2021/11/17/University-of-PIttsburgh-CMU-hospital-infection-outbreaks-artificial-intelligence/stories/202111160095
> but we should probably expect more noise.
>
> This is another good example of how AI research done at the Auton Lab
> can make a tangible impact in the world.
>
> Way to go Auton EDS-HAT Team! (it prominently includes Kyle and Jessie
> all the time from the start of the project some six years ago, and
> most recently adds Dan, Andrew and Stef)
>
> Cheers,
> Artur
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Pittsburgh Hospital Taps AI to Prevent Spread of Infections.pdf
Type: application/pdf
Size: 314997 bytes
Desc: not available
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20211124/81515a87/attachment-0001.pdf>

From awd at cs.cmu.edu  Mon Nov 29 16:47:14 2021
From: awd at cs.cmu.edu (Artur Dubrawski)
Date: Mon, 29 Nov 2021 16:47:14 -0500
Subject: Ben's thesis proposal this Wednesday at 4pm
Message-ID: <CAJvAoyviMwt90aG_vqRLzH7W3mXkyGEHAgiXHdf5sKaAkfe1jg@mail.gmail.com>

Team,

Please consider joining Benedikt Boecking's thesis proposal
presentation - which will be on zoom only - this Wednesday December
1st at 4pm.

Details below.

Cheers,
Artur

Title:
Learning with Diverse Forms of Imperfect and Indirect Supervision

Abstract:
High capacity Machine Learning (ML) models trained on large, annotated
datasets have driven impressive advances in several fields including
natural language processing and computer vision, in turn leading to
impactful applications of ML in areas such as healthcare, e-commerce,
and predictive maintenance. However, obtaining annotated datasets at
the scale required for training such models is costly and often
becomes a bottleneck for promising applications of ML. In this thesis,
I study imperfect and indirect forms of supervision (weak supervision)
such as partial rules and pairwise constraints as a mechanism to
encode domain knowledge, as these are frequently easy to obtain at
scale and can enable learning without pointillistic ground truth
annotations.

I begin by studying the utility of small amounts of pairwise
supervision for clustering, by using known group-membership
constraints to learn a kernel to improve constrained clustering
performance. Next, I propose a methodology that uses imperfect
pairwise labels to augment learning for programmatic data labeling
methods which traditionally only learn from Labeling Functions (LFs),
i.e. user defined functions that directly but imperfectly label
subsets of data. Such label models aggregate sources of imperfect
supervision to estimate the latent ground truth and act as teachers to
end models, thereby playing an essential role in achieving
generalization. Preliminary results show promising performance
improvements.

I further the study of programmatic data labeling methods by
introducing integrated, end-to-end learning frameworks and novel label
models. I first introduce a framework for joint learning of a label
and end models from LFs, showing improved performance over prior work
in terms of end model performance on downstream test sets. I then
propose a new methodology based on discrete latent variable modeling
in generative adversarial networks to improve estimates of the
unobserved ground truth through uncovering of disentangled, discrete
structures in the features.

Finally, I study two extremes on the spectrum of domain knowledge
acquisition in weak supervision: user interactivity for discovering
useful sources of imperfect labels, and learning merely from data
paired with unstructured natural language descriptions. I first
introduce an interactive learning framework that aids users in
discovering weak supervision sources to systematically and proactively
capture subject matter experts? knowledge of the application domain in
an efficient and effective fashion. I then propose to study how
unstructured natural language descriptions (such as doctors notes)
paired with images can be exploited for image representation learning
and zero-shot classification, without requiring experts to define
rules on the text or images as in prior related work.

Together, these works provide novel methodologies and frameworks to
more efficiently encode expert domain knowledge in ML models, reducing
the bottleneck created by the need for pointillistic ground truth
annotations.

Thesis Committee
Artur Dubrawski (Chair)
Barnab?s P?czos
Jeff Schneider
Hoifung Poon (Microsoft Research)

Zoom Meeting ID: 936 5393 2784 Passcode: 877794