From chiragn at cs.cmu.edu  Tue Jul  2 17:52:04 2019
From: chiragn at cs.cmu.edu (Chirag Nagpal)
Date: Tue, 2 Jul 2019 17:52:04 -0400
Subject: Harvard Data Science Review
Message-ID: <CAKH1gVqnXS4DAa47-FeOGsm=CTtRpWD-5UEJ2TE+70+J36JX7g@mail.gmail.com>

Harvard and MIT have jointly launched a "Review Journal" called Data
Science Review. Something on the lines of MIT Tech Review/Harvard Business
Review but for Data Science.

The link is below
https://hdsr.mitpress.mit.edu/

Might be of interest to many members in lab.

Chirag


-- 

*Chirag Nagpal* Graduate Student, Language Technologies Institute
School of Computer Science
Carnegie Mellon University
cs.cmu.edu/~chiragn
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190702/7cbc9f4b/attachment.html>

From predragp at andrew.cmu.edu  Sat Jul  6 15:59:45 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Sat, 06 Jul 2019 15:59:45 -0400
Subject: file server problems
Message-ID: <20190706195945.ffn3e5Krw%predragp@andrew.cmu.edu>

Dear Autonians,

I hate to interrupt your 4th of July festivities with a bad news but it
appears that I have a major problem with one of our main file servers
after a minor upgrade. This server is the home to about 2/3 of the Auton
Lab home directories which means that many of you are no longer able to
log into our computing infrastructure apart of your Auton Lab maintained
desktops. 

Before anyone gets too upset at me for upgrading file server let me just
state for the record that the upgrade was tested and went smoothly 
on 6 other machines. As you know upgrading software is a natural part of
the machine's life cycle. 

At this point I am unable to cleanly boot machine which is stuck while
trying to mount the root file system. I am planning to boot machine with
the live media first thing on Monday morning and if that doesn't work
just reinstall OS and import ZFS pools containing your home directories
(they should not be affected). 

Most Kind Regards,
Predrag Punosevac

P.S. Dr. Jeff Schneider has purchased a new file server which would have
been a backup machine in this case. Unfortunately we have not received
it yet so I will have to troubleshoot this.

From predragp at andrew.cmu.edu  Sun Jul  7 11:50:51 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Sun, 7 Jul 2019 11:50:51 -0400
Subject: file server problems
In-Reply-To: <20190706195945.ffn3e5Krw%predragp@andrew.cmu.edu>
References: <20190706195945.ffn3e5Krw%predragp@andrew.cmu.edu>
Message-ID: <CAEFFt-FW2KVu+-gebvpEUNCufdXicK8o+3RKNch2os0sORieiA@mail.gmail.com>

Dear Autonians,

I am happy to report that gaia is up and running so all the users with
home directories
on zfsauton2 and zfsauton3 can resume their work. However, I am not
happy that I had manually (in person)
to boot the machine. I am frantically checking the hardware
and going through log files trying to figure out what has happened.


Best,
Predrag


On Sat, Jul 6, 2019 at 3:59 PM Predrag Punosevac
<predragp at andrew.cmu.edu> wrote:
>
> Dear Autonians,
>
> I hate to interrupt your 4th of July festivities with a bad news but it
> appears that I have a major problem with one of our main file servers
> after a minor upgrade. This server is the home to about 2/3 of the Auton
> Lab home directories which means that many of you are no longer able to
> log into our computing infrastructure apart of your Auton Lab maintained
> desktops.
>
> Before anyone gets too upset at me for upgrading file server let me just
> state for the record that the upgrade was tested and went smoothly
> on 6 other machines. As you know upgrading software is a natural part of
> the machine's life cycle.
>
> At this point I am unable to cleanly boot machine which is stuck while
> trying to mount the root file system. I am planning to boot machine with
> the live media first thing on Monday morning and if that doesn't work
> just reinstall OS and import ZFS pools containing your home directories
> (they should not be affected).
>
> Most Kind Regards,
> Predrag Punosevac
>
> P.S. Dr. Jeff Schneider has purchased a new file server which would have
> been a backup machine in this case. Unfortunately we have not received
> it yet so I will have to troubleshoot this.
>

From awertz at cmu.edu  Mon Jul  8 11:23:09 2019
From: awertz at cmu.edu (Anthony Wertz)
Date: Mon, 8 Jul 2019 11:23:09 -0400
Subject: lov3 scratch full; please consider removing unneeded files
Message-ID: <CACOCF0JvgaSFP1GdT3V1-z=KbCaeUY-6mqNWf_8saDcOhcSdTQ@mail.gmail.com>

All,

lov3 scratch is full. If you have some results you no longer need or can
store elsewhere, we would greatly appreciate freeing up some space so we
can continue working there.

-- 
*Anthony Wertz*
Research Programmer and Analyst
Robotics Institute - Auton Lab
Carnegie Mellon University
awertz at cmu.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190708/ea1bea98/attachment.html>

From predragp at andrew.cmu.edu  Mon Jul  8 12:14:36 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Mon, 8 Jul 2019 12:14:36 -0400
Subject: SSH Problems with LOP1
In-Reply-To: <CAEFFt-EtFh++8BtzRZmWHvNOetheZiOM04pb4VZZT155JaiEHg@mail.gmail.com>
References: <CAKH1gVopMuUATLJgPq2sQF5WUuHcRoPifjVv6SW=iv6hWkMkpQ@mail.gmail.com>
 <CAKH1gVrutsgnobKJ5hq-cC2Rrh41aGfatcj0KZFs0jk_vfaaBg@mail.gmail.com>
 <CAKH1gVpTOXayaLbLZvuNbzww+X6NwxnchSGkrum9CjqQWU66Ug@mail.gmail.com>
 <20190708021923.d_Lq7JEmB%predragp@andrew.cmu.edu>
 <CAKH1gVqjFj-EL9ORFj7PWpWX3pRi5y-YNA8UoNszJ6EYJ+Q61g@mail.gmail.com>
 <CAEFFt-G=5FOOc2CGJHK1VTS-Cvk-TQ0Y3d4HSDUOMvZtjHzyFA@mail.gmail.com>
 <CAKH1gVqQB1hNYbe2mmkn3kjT7bW-3QHkdCoAHoLMca4B4edKDw@mail.gmail.com>
 <CAEFFt-EtFh++8BtzRZmWHvNOetheZiOM04pb4VZZT155JaiEHg@mail.gmail.com>
Message-ID: <CAEFFt-FCEKEHUQPGdu=DGheH=HB8tnRgSPRzKrVkLXYZf_jzxg@mail.gmail.com>

This is legit! I can confirm this as well as Rob. For some reason
ypldap deamon is failing to start. I just restart manually and was
able to log both from bash but more importantly from my home. That
machine was due for replacement. Please give me a day or two to
provision a new machine. In the meantime you are stuck with bash and
lop1 whenever is available (right now it is).

Predrag

On Mon, Jul 8, 2019 at 9:24 AM Chirag Nagpal <chiragn at cs.cmu.edu> wrote:
>
> I am unable to ssh to lop1, I get connection terminated by remote host.
>
> I tried sshing into lop1 from bash, which is also timing out.
>
>
>

From predragp at andrew.cmu.edu  Tue Jul  9 16:41:21 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Tue, 9 Jul 2019 16:41:21 -0400
Subject: CUDA out of memory
In-Reply-To: <CADpAR=1OT_zpgmuj0ovuuFnBVphZZ11P2dJ58TwLnuuEzPK3CA@mail.gmail.com>
References: <CADpAR=1OT_zpgmuj0ovuuFnBVphZZ11P2dJ58TwLnuuEzPK3CA@mail.gmail.com>
Message-ID: <CAEFFt-GhoaJOEzasLA7yFC=GoW-3n7wC-M4fSBaqgiwOAx2+3w@mail.gmail.com>

Send an email to users at autonlab.org and ask politely people to free up
memory. Locking up memory is a "feature" of deep learning software.

On Tue, Jul 9, 2019 at 3:38 PM Shuli Jiang <shulij at alumni.cmu.edu> wrote:
>
> Hello, Predrag,
>
> I found lab gpu machines have a problem today: cuda out of memory.
> Could you please take a look?
> Thanks a lot.
>
> Best,
> Shuli

From chiragn at cs.cmu.edu  Tue Jul  9 16:49:45 2019
From: chiragn at cs.cmu.edu (Chirag Nagpal)
Date: Tue, 9 Jul 2019 16:49:45 -0400
Subject: CUDA out of memory
In-Reply-To: <CAEFFt-GhoaJOEzasLA7yFC=GoW-3n7wC-M4fSBaqgiwOAx2+3w@mail.gmail.com>
References: <CADpAR=1OT_zpgmuj0ovuuFnBVphZZ11P2dJ58TwLnuuEzPK3CA@mail.gmail.com>
 <CAEFFt-GhoaJOEzasLA7yFC=GoW-3n7wC-M4fSBaqgiwOAx2+3w@mail.gmail.com>
Message-ID: <CAKH1gVr-F6YYy27ZTNj98qtQjfvovgMANE6YOGGe669TM9DrJw@mail.gmail.com>

Hi Shuli

Before using a particular machine, please use nvidia-smi to see if the
machine you are using has a card with free Video RAM.

Once you are sure of the machine and the card you want to use please use

$EXPORT CUDA_VISIBLE_DEVICES=k

on the machine  with k being the free GPU Card.

this would ensure your CUDA job is only resident in the target GPU.

Best,

Chirag

On Tue, Jul 9, 2019 at 4:46 PM Predrag Punosevac <predragp at andrew.cmu.edu>
wrote:

> Send an email to users at autonlab.org and ask politely people to free up
> memory. Locking up memory is a "feature" of deep learning software.
>
> On Tue, Jul 9, 2019 at 3:38 PM Shuli Jiang <shulij at alumni.cmu.edu> wrote:
> >
> > Hello, Predrag,
> >
> > I found lab gpu machines have a problem today: cuda out of memory.
> > Could you please take a look?
> > Thanks a lot.
> >
> > Best,
> > Shuli
>


-- 

*Chirag Nagpal* Graduate Student, Language Technologies Institute
School of Computer Science
Carnegie Mellon University
cs.cmu.edu/~chiragn
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190709/e92fd8e0/attachment.html>

From shulij at alumni.cmu.edu  Tue Jul  9 18:18:24 2019
From: shulij at alumni.cmu.edu (Shuli Jiang)
Date: Tue, 9 Jul 2019 18:18:24 -0400
Subject: CUDA out of memory
In-Reply-To: <CAKH1gVr-F6YYy27ZTNj98qtQjfvovgMANE6YOGGe669TM9DrJw@mail.gmail.com>
References: <CADpAR=1OT_zpgmuj0ovuuFnBVphZZ11P2dJ58TwLnuuEzPK3CA@mail.gmail.com>
 <CAEFFt-GhoaJOEzasLA7yFC=GoW-3n7wC-M4fSBaqgiwOAx2+3w@mail.gmail.com>
 <CAKH1gVr-F6YYy27ZTNj98qtQjfvovgMANE6YOGGe669TM9DrJw@mail.gmail.com>
Message-ID: <CADpAR=0eywFCD+btfm1hDH6b7BeDFjvm6yr-gjm4A119pn05KA@mail.gmail.com>

Thank you Chirag. It solves the problem now.

On Tue, Jul 9, 2019 at 4:49 PM Chirag Nagpal <chiragn at cs.cmu.edu> wrote:

> Hi Shuli
>
> Before using a particular machine, please use nvidia-smi to see if the
> machine you are using has a card with free Video RAM.
>
> Once you are sure of the machine and the card you want to use please use
>
> $EXPORT CUDA_VISIBLE_DEVICES=k
>
> on the machine  with k being the free GPU Card.
>
> this would ensure your CUDA job is only resident in the target GPU.
>
> Best,
>
> Chirag
>
> On Tue, Jul 9, 2019 at 4:46 PM Predrag Punosevac <predragp at andrew.cmu.edu>
> wrote:
>
>> Send an email to users at autonlab.org and ask politely people to free up
>> memory. Locking up memory is a "feature" of deep learning software.
>>
>> On Tue, Jul 9, 2019 at 3:38 PM Shuli Jiang <shulij at alumni.cmu.edu> wrote:
>> >
>> > Hello, Predrag,
>> >
>> > I found lab gpu machines have a problem today: cuda out of memory.
>> > Could you please take a look?
>> > Thanks a lot.
>> >
>> > Best,
>> > Shuli
>>
>
>
> --
>
> *Chirag Nagpal* Graduate Student, Language Technologies Institute
> School of Computer Science
> Carnegie Mellon University
> cs.cmu.edu/~chiragn
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190709/f81ebee7/attachment.html>

From ngisolfi at cs.cmu.edu  Thu Jul 11 08:23:17 2019
From: ngisolfi at cs.cmu.edu (Nick Gisolfi)
Date: Thu, 11 Jul 2019 08:23:17 -0400
Subject: [Lunch] Today @12-1pm in NSH 4201
Message-ID: <B376858B-9CBD-416E-AE2D-9BD490D7B384@cs.cmu.edu>

Hi Everyone,

Lab lunch will be in NSH 4201 from 12-1pm today.

Bring lunch and we?ll see you there!

As always, please invite anyone who may not be on the mailing list but still working with us this summer.

- Nick

From awd at cs.cmu.edu  Thu Jul 11 14:59:46 2019
From: awd at cs.cmu.edu (Artur Dubrawski)
Date: Thu, 11 Jul 2019 14:59:46 -0400
Subject: Fwd: FW: HCII Special Seminar Speaker: James Landay (Monday, July 22)
In-Reply-To: <fe45b30d3a354b3d991c5a13619800bd@cs.cmu.edu>
References: <EC76EA7C-D173-4F08-A34B-0F29794D699A@andrew.cmu.edu>
 <ca8e21b0-9572-3910-6457-21ceb6761b12@cs.cmu.edu>
 <fe45b30d3a354b3d991c5a13619800bd@cs.cmu.edu>
Message-ID: <CAJvAoyt4mueKp1TEXZE37YE4C1R7vedzsGg-CNrf3HTOvr-9LQ@mail.gmail.com>

Many of us are involved in relevant work so i thought i would broadcast
this to avoid missing anyone actually interested.

Cheers
Artur

---------- Forwarded message ---------
From: Roni Rosenfeld <roni at cs.cmu.edu>
Date: Thu, Jul 11, 2019, 2:56 PM
Subject: FW: HCII Special Seminar Speaker: James Landay (Monday, July 22)
To: ML-faculty <ml-faculty at cs.cmu.edu>


Topic may be of interest to ML folks.  He's also helping to push the HCI/AI
initiative at Stanford.

-------- Forwarded Message --------
Subject:        HCII Special Seminar Speaker: James Landay (Monday, July 22)
Date:   Thu, 11 Jul 2019 13:53:54 -0400
From:   Carolyn Buzzelli-Stumpf <cbstumpf at andrew.cmu.edu>
To:     hcii-faculty-mtgs at cs.cmu.edu

Dear Faculty,

HCII Special Seminar Speaker, James Landay will be here on Monday, July 22.
HIs talk is scheduled at 12:00 - 1:00pm in Newell Simon Hall, Room 4305.

If you would like to meet with James on Monday, 7/22,  please sign up
here: James Landay schedule: July 22
<
https://docs.google.com/spreadsheets/d/1kG4RkLJeFGQRaDER1BTmtphDpysSKN4bI_3duBbNT9g/edit?usp=sharing
>


*Smart Interfaces for Human-Centered AI*

Professor James A. Landay
Computer Science Department
Stanford University

*Abstract*

AI has the potential to automate people out of their jobs, and in some
cases, it will. But while we should carefully consider the risk of
replacing human capabilities, it?s important to realize that AI has
enormous potential to /augment/ them as well: it can boost the creativity
of our work, help us learn better, deliver healthcare more effectively, and
make our societies more sustainable. Like any tool, however, our
relationship with AI has as much to do with its interface as it does with
the underlying capabilities it provides. Does it amplify our actions and
remain attentive to our goals?even as we revise them?or is it a black box
that accomplishes tasks autonomously? If we want to build a future of open
possibility and empowerment, it?s vital that our ability to harness AI
evolves alongside AI itself. I will illustrate how we are addressing grand
challenges in fields like health and education by building systems that
balance innovative user interfaces with intell!
 igent systems.

*Bio*

James Landay is a Professor of Computer Science and the Anand Rajaraman and
Venky Harinarayan Professor in the School of Engineering at Stanford
University. He specializes in human-computer interaction. He is the founder
and co-director of the World Lab, a joint research and educational effort
with Tsinghua University in Beijing. He is also the Associate Director of
the Stanford Institute for Human-centered Artificial Intelligence (HAI).
Previously, Landay was a Professor of Information Science at Cornell Tech
in New York City and prior to that he was a Professor of Computer Science &
Engineering at the University of Washington. From 2003 through 2006 he was
the Laboratory Director of Intel Labs Seattle, a university affiliated
research lab that explored the new usage models, applications, and
technology for ubiquitous computing. He was also the chief scientist and
co-founder of NetRaker, which was acquired by KeyNote Systems in 2004. From
1997 through 2003 he was a professo!
 r in EECS at UC Berkeley. Landay received his BS in EECS from UC Berkeley
in 1990, and MS and PhD in Computer Science from Carnegie Mellon University
in 1993 and 1996, respectively. His PhD dissertation was the first to
demonstrate the use of sketching in user interface design tools. He was
named to the ACM SIGCHI Academy in 2011 and as an ACM Fellow in 2017. He
formerly served on the NSF CISE Advisory Committee.


Thank you,
Carolyn


*Carolyn Buzzelli-Stumpf* | Assistant to the Director | Human-Computer
Interaction Institute** *The School of Computer Science*| *Carnegie Mellon
University * 3521Newell-Simon Hall | Pittsburgh, PA 15213 | 412-268-1001
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190711/fc854bac/attachment.html>

From predragp at andrew.cmu.edu  Wed Jul 17 18:16:54 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Wed, 17 Jul 2019 18:16:54 -0400
Subject: Main File server RAM upgrade
Message-ID: <CAEFFt-FXP7cREDwOVQ4UddmO6dReJNwNnEaAm1DeAzOREMdQ8Q@mail.gmail.com>

Dear Autonians,

I just beefed-up RAM (paid by Dr. Dubrawski) on the main file server
from 64 GB to 128 GB. If your home directories are located on
zfsauton2 or zfsauton3 you have notice a 10 minute blip while machine
was powered off. I apologize for the inconvenience but many people are
away from Pittsburgh and this is a good time to do this kind
disruptive maintenance.

Cheers,
Predrag

From ngisolfi at cs.cmu.edu  Thu Jul 18 08:27:30 2019
From: ngisolfi at cs.cmu.edu (Nick Gisolfi)
Date: Thu, 18 Jul 2019 08:27:30 -0400
Subject: [Lunch] Today @12-1pm in NSH 4201
Message-ID: <32266B7E-06E4-4A20-ACD0-4A9C8A9B4F55@cs.cmu.edu>

Hi Everyone,

Lab lunch will be in NSH 4201 today from 12-1pm.

Bring your lunch and we?ll see you there!

- Nick

From predragp at andrew.cmu.edu  Wed Jul 24 13:09:56 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Wed, 24 Jul 2019 13:09:56 -0400
Subject: No GPU drivers detected on any gpu machine?
In-Reply-To: <b281b6da8c5c4a998701a82a3e1d0ed5@andrew.cmu.edu>
References: <b281b6da8c5c4a998701a82a3e1d0ed5@andrew.cmu.edu>
Message-ID: <CAEFFt-FQm3d2Km59cXUS0miLLjVSXOUy090h=TPr71tQExdykA@mail.gmail.com>

Thank you so much for bringing this to my attention. GPU10 is not
broken but sure enough you are right about the other machines. It
appears that one of recent updates have broken the driver. I will
reinstall drivers shortly and reboot the machines. This is also notice
for everyone else that GPU1-9 will have to be rebooted.

Predrag

On Wed, Jul 24, 2019 at 10:52 AM Chufan Gao <chufang at andrew.cmu.edu> wrote:
>
> Hi Predrag,
>
>
> I discovered today that when I run nvidia-smi, I get this error:
>
>
> NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
>
> The same happens for all of the gpu machines that I tried. I am confused - was there an update that broke it?
>
> Sincerely,
> Andy Gao

From predragp at andrew.cmu.edu  Wed Jul 24 16:15:22 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Wed, 24 Jul 2019 16:15:22 -0400
Subject: No GPU drivers detected on any gpu machine?
In-Reply-To: <CAEFFt-FQm3d2Km59cXUS0miLLjVSXOUy090h=TPr71tQExdykA@mail.gmail.com>
References: <b281b6da8c5c4a998701a82a3e1d0ed5@andrew.cmu.edu>
 <CAEFFt-FQm3d2Km59cXUS0miLLjVSXOUy090h=TPr71tQExdykA@mail.gmail.com>
Message-ID: <CAEFFt-Fk69-0vs=cg4KGvORKZq2hgJ-TRwCbKQsMyhsQy4Aj0Q@mail.gmail.com>

A quick update on this issue and a resolution. I took a clue from the
fact that GPU10 was working as expected and narrowed down the issue to
CUDA 9.1 installation.  It appears that upstream has broken CUDA 9.1
purposely via dkms utility which is used to recompile kernel modules
to fit specific kernel release. They probably want people to move to
CUDA 10.1.

Long story short. I upgraded NVidia driver and CUDA to 10.1 on GPU2
and GPU3 servers. They appear to be working flawlessly on my end as
tested with nvidia-smi utility as well as MATLAB. I have recreated
GPU3 scratch directory which was 100% used for almost half a year. I
have also reinstalled libcudnn library on both machines but I am
unable to test it.

This is all good but it also means that people will have to regenerate
their tools from the scratch to match the kernel, driver, and CUDA
versions. If you have things on GPU10 you probably could just migrate
them. This is very time consuming but we have no choice.

The major bad news is that one of the GPU servers I tried to work on
GPU1 (commissioned almost five years ago) didn't survive reboot. It
also uses older Tesla K80 cards. I will have to attach the screen and
troubleshoot this machine. That will not happen today or for that
matter this week.

My plan is now to move and fix machines GPU[4-9] which would take the
rest of the day.Note that GPU7 is designated for a special project and
not generally accessible.

Most Kind Regards,
Predrag Punosevac




On Wed, Jul 24, 2019 at 1:09 PM Predrag Punosevac
<predragp at andrew.cmu.edu> wrote:
>
> Thank you so much for bringing this to my attention. GPU10 is not
> broken but sure enough you are right about the other machines. It
> appears that one of recent updates have broken the driver. I will
> reinstall drivers shortly and reboot the machines. This is also notice
> for everyone else that GPU1-9 will have to be rebooted.
>
> Predrag
>
> On Wed, Jul 24, 2019 at 10:52 AM Chufan Gao <chufang at andrew.cmu.edu> wrote:
> >
> > Hi Predrag,
> >
> >
> > I discovered today that when I run nvidia-smi, I get this error:
> >
> >
> > NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
> >
> > The same happens for all of the gpu machines that I tried. I am confused - was there an update that broke it?
> >
> > Sincerely,
> > Andy Gao

From predragp at andrew.cmu.edu  Wed Jul 24 22:13:28 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Wed, 24 Jul 2019 22:13:28 -0400
Subject: No GPU drivers detected on any gpu machine?
In-Reply-To: <CAEFFt-Fk69-0vs=cg4KGvORKZq2hgJ-TRwCbKQsMyhsQy4Aj0Q@mail.gmail.com>
References: <b281b6da8c5c4a998701a82a3e1d0ed5@andrew.cmu.edu>
 <CAEFFt-FQm3d2Km59cXUS0miLLjVSXOUy090h=TPr71tQExdykA@mail.gmail.com>
 <CAEFFt-Fk69-0vs=cg4KGvORKZq2hgJ-TRwCbKQsMyhsQy4Aj0Q@mail.gmail.com>
Message-ID: <20190725021328.b1FI1L2xq%predragp@andrew.cmu.edu>

Predrag Punosevac <predragp at andrew.cmu.edu> wrote:

I apologize for top posting. Just a quick update. As of 5 minutes ago
machines gpu[2-10] appear to have no issues. After all the upgrades and
reboots it appears that we don't have any dead GPU cards on them and
that drivers and CUDA 10.1 work as expected. I understand that this is a
little comfort to people who need to regenerate tensorflow, py-torch,
and all that "deep-learning" stuff but I have no control over the
upstream decisions. 

GPU1 appears to be broken at the moment. Without attaching consol to the
machine it is difficult for me to asses the complexity of the problem. 

One more time sorry for the down time. 

Cheers,
Predrag






> A quick update on this issue and a resolution. I took a clue from the
> fact that GPU10 was working as expected and narrowed down the issue to
> CUDA 9.1 installation.  It appears that upstream has broken CUDA 9.1
> purposely via dkms utility which is used to recompile kernel modules
> to fit specific kernel release. They probably want people to move to
> CUDA 10.1.
> 
> Long story short. I upgraded NVidia driver and CUDA to 10.1 on GPU2
> and GPU3 servers. They appear to be working flawlessly on my end as
> tested with nvidia-smi utility as well as MATLAB. I have recreated
> GPU3 scratch directory which was 100% used for almost half a year. I
> have also reinstalled libcudnn library on both machines but I am
> unable to test it.
> 
> This is all good but it also means that people will have to regenerate
> their tools from the scratch to match the kernel, driver, and CUDA
> versions. If you have things on GPU10 you probably could just migrate
> them. This is very time consuming but we have no choice.
> 
> The major bad news is that one of the GPU servers I tried to work on
> GPU1 (commissioned almost five years ago) didn't survive reboot. It
> also uses older Tesla K80 cards. I will have to attach the screen and
> troubleshoot this machine. That will not happen today or for that
> matter this week.
> 
> My plan is now to move and fix machines GPU[4-9] which would take the
> rest of the day.Note that GPU7 is designated for a special project and
> not generally accessible.
> 
> Most Kind Regards,
> Predrag Punosevac
> 
> 
> 
> 
> On Wed, Jul 24, 2019 at 1:09 PM Predrag Punosevac
> <predragp at andrew.cmu.edu> wrote:
> >
> > Thank you so much for bringing this to my attention. GPU10 is not
> > broken but sure enough you are right about the other machines. It
> > appears that one of recent updates have broken the driver. I will
> > reinstall drivers shortly and reboot the machines. This is also notice
> > for everyone else that GPU1-9 will have to be rebooted.
> >
> > Predrag
> >
> > On Wed, Jul 24, 2019 at 10:52 AM Chufan Gao <chufang at andrew.cmu.edu> wrote:
> > >
> > > Hi Predrag,
> > >
> > >
> > > I discovered today that when I run nvidia-smi, I get this error:
> > >
> > >
> > > NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
> > >
> > > The same happens for all of the gpu machines that I tried. I am confused - was there an update that broke it?
> > >
> > > Sincerely,
> > > Andy Gao

From ngisolfi at cs.cmu.edu  Thu Jul 25 08:28:00 2019
From: ngisolfi at cs.cmu.edu (Nick Gisolfi)
Date: Thu, 25 Jul 2019 08:28:00 -0400
Subject: [Lunch] Today @ 12-1pm in NSH 4201
Message-ID: <7992B40E-93FC-4267-8B75-FE83D311CDF3@cs.cmu.edu>

Hi Everyone,

Lab lunch will be in NSH 4201 from 12-1pm today.

Bring your meal and we?ll see you there!

- Nick

From predragp at andrew.cmu.edu  Thu Jul 25 10:26:57 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Thu, 25 Jul 2019 10:26:57 -0400
Subject: updates
Message-ID: <20190725142657.GhqxxD9Iv%predragp@andrew.cmu.edu>

Dear Autonians,

Just a quick update. This morning I patched all our network machines
(Auton Sytems and NREC included) for TCP SACK holes per

https://www.openbsd.org/errata65.html

I also upgrade all file servers (again Auton Systems and NREC included)
and FreeBSD jails to

11.3-RELEASE-p1

Everything seems to be working as expected. This might have been
observed by some of you as a short interruption in the traffic. Sorry
for that. Machine with a very long uptimes are by definition
unmaintained.

Predrag

From ngisolfi at cs.cmu.edu  Thu Jul 25 11:36:09 2019
From: ngisolfi at cs.cmu.edu (Nick Gisolfi)
Date: Thu, 25 Jul 2019 11:36:09 -0400
Subject: UPDATE [Lunch] Today @ 12-1pm OUTSIDE
In-Reply-To: <7992B40E-93FC-4267-8B75-FE83D311CDF3@cs.cmu.edu>
References: <7992B40E-93FC-4267-8B75-FE83D311CDF3@cs.cmu.edu>
Message-ID: <37500295-3A18-428C-9843-E1303BBAB193@cs.cmu.edu>

On second thought, it sounds like some of us like the idea of having lab lunch outside.  Let?s try to meet on the 6th floor Gates balcony.  Hopefully we can push together enough tables and the overhang provides enough shade.

- Nick


> On Jul 25, 2019, at 8:28 AM, Nick Gisolfi <ngisolfi at cs.cmu.edu> wrote:
> 
> Hi Everyone,
> 
> Lab lunch will be in NSH 4201 from 12-1pm today.
> 
> Bring your meal and we?ll see you there!
> 
> - Nick



From predragp at andrew.cmu.edu  Fri Jul 26 14:46:31 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Fri, 26 Jul 2019 14:46:31 -0400
Subject: Performance issue with zfsauton?
In-Reply-To: <CAAgoO26eq=+k4DDeTgCv8y7U12vyW2R5XPk-JPNhXMY8yO7-uw@mail.gmail.com>
References: <CAAgoO26eq=+k4DDeTgCv8y7U12vyW2R5XPk-JPNhXMY8yO7-uw@mail.gmail.com>
Message-ID: <CAEFFt-GSgvNEQPdBv_cdXub6Jt49KZhvAMpFUgw+T_nVYxs+7g@mail.gmail.com>

Not to zfsauton. I added more memory to gaia i.e zfsauton2 and
zfsauton3. You are the second person reporting this so I will have to
fire up DTrace to see what is happening.

Predrag

On Fri, Jul 26, 2019 at 2:33 PM Robert MacLachlan <robmacl at cmu.edu> wrote:
>
> I notice you added more memory to the server, but I am still see lots of jobs that used to be CPU bound bogging down, and in at least some cases I have seen them to be blocked trying to read from /zfsauton/data.  This seems to be a new thing.  Is someone pushing a super-lot of data into/out of zfs? Well, my bad for doing so myself, but...
>
>   Rob
>


From mdeartea at andrew.cmu.edu  Sat Jul 27 11:23:42 2019
From: mdeartea at andrew.cmu.edu (Maria De Arteaga Gonzalez)
Date: Sat, 27 Jul 2019 11:23:42 -0400
Subject: Call for papers: 3rd NeurIPS workshop on ML for the Developing World
 (ML4D)
Message-ID: <CACsvUsbCtJNW1fW1jYY13qi+gQ6nc=hxCMzeuxvTgNzLYzCtOw@mail.gmail.com>

Hi all,

We are very excited to announce the 3rd NeurIPS workshop on Machine
Learning for the Developing World (ML4D), please see the CFP below. We'd
also appreciate your help spreading the word. Thanks!

Maria

Call for papers: NeurIPS workshop on ML for the Developing World (ML4D)

*********************************************************************************

Workshop on Machine Learning for the Developing World, NeurIPS 2019

Date: December 13th or 14th (TBD), 2019

Location: Vancouver, Canada

Website: https://sites.google.com/view/ml4d/home

*********************************************************************************

Call for papers:

For the third year in a row, NeurIPS is host to a one-day workshop focussed
on machine learning for the developing world (ML4D). This year?s program
will focus on the challenges and risks that arise when deploying machine
learning in developing regions.


We invite researchers to submit their recent work on ML4D, and particularly
encourage submissions that characterize challenges or risks of ML4D,
propose methodology tackling existing limitations, present empirical
studies that reveal unintended harms of machine learning technology in
developing regions, or discuss the current state of the art and propose
paths forward.

Please submit 2-4 page extended abstracts following the NeurIPS style
<https://nips.cc/Conferences/2017/PaperInformation/StyleFiles> guidelines.
The link to the submission platform will be provided in the workshop
website. Accepted papers will be presented as posters or contributed talks,
and may opt-in to be published in an arXiv proceedings.


Key dates:

Submission deadline: September 13, 2019

Travel/registration award deadline: September 13, 2019

Acceptance notification: September 25, 2019

Workshop: December 13/14 (TBD), 2019


Workshop overview:


As the use of machine learning becomes ubiquitous, there is growing
interest in understanding how machine learning can be used to tackle global
development challenges. The possibilities are vast, and it is important
that we explore the potential benefits of such technologies, which has
driven the agenda of the ML4D workshop series in the past. However, there
is a risk that technology optimism and a categorization of ML4D research as
inherently ?social good? may result in initiatives failing to account for
unintended harms or deviating scarce funds towards initiatives that appear
exciting but have no demonstrated effect. Moreover, machine learning
technologies deployed in developing regions have often been created for
different contexts and are trained with data that is not representative of
the new deployment setting. Most concerning of all, multinational companies
sometimes make the deliberate choice to deploy new technologies in
countries with little regulation in order to experiment. This year?s
program will focus on the challenges and risks that arise when deploying
machine learning in developing regions. This one-day workshop will bring
together a diverse set of participants from across the globe to discuss
essential elements for ensuring ML4D research moves forward in a
responsible and ethical manner. Attendees will learn about potential
unintended harms that may result from ML4D solutions, technical challenges
that currently prevent the effective use of machine learning in vast
regions of the world, and lessons that may be learned from other fields.
The workshop will include invited talks, a poster session of accepted
papers, breakout sessions tailored to the workshop's theme and panel
discussions. We welcome paper submissions featuring novel machine learning
research that characterizes or tackles challenges of ML4D, empirical papers
that reveal unintended harms of machine learning technology in developing
regions, and discussion papers that examine the current state of the art of
ML4D and propose paths forward.


Maria De-Arteaga
PhD Student in Machine Learning and Public Policy
Carnegie Mellon University
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190727/ac0ebe6f/attachment.html>

From predragp at andrew.cmu.edu  Sat Jul 27 13:48:36 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Sat, 27 Jul 2019 13:48:36 -0400
Subject: Weird networking problem on home net access to CMU
In-Reply-To: <CAAgoO27kAGbH+yrQWaB8KHTyiLNVpyrYr6g8sTnKB+6Js7nrpA@mail.gmail.com>
References: <CAAgoO27kAGbH+yrQWaB8KHTyiLNVpyrYr6g8sTnKB+6Js7nrpA@mail.gmail.com>
Message-ID: <20190727174836.VkKkDCF_F%predragp@andrew.cmu.edu>

Robert MacLachlan <robmacl at cmu.edu> wrote:

> I have verizon FIOS and having this weird problem where I can access
> anything in the entire world except the CMU and auton nets.  Is there
> any technical way to diagnose where the problem is, i.e. am I some how
> blocked off of the CMU net by CMU, or is there some choke point in the
> Verizon net, or what?
> 
> Does CMU have an IP address blocklist?
> 
> I am connecting by tethering to my phone (T-moble) which has no
> problem, but this is of course slow.  The problem started this week,
> was there on monday, then stopped for a while and came back maybe
> thursday.  At first I thought that the CMU net was actually down.
> 
>   Rob

Hi Rob,

I have seen and experienced this first hand myself few years ago with
Armstrong cable as my ISP. I went to a great deal of network
troubleshooting (lot of traceroute, dig, tcpdump, taking to CMU network
guys including off site) without getting conclusive evidence pointing to
any particular reason.

For starters you can try to traceroute www.cmu.edu from your machine as
well as from one of content global delivery networks. 

https://tools.keycdn.com/traceroute

You will need to use dig and whois to convert all those IP addresses to
domain names and legal entities.  I can see that at this very moment CMU
network is not reachable from Miami and San Francisco servers.
Unfortunately the breaking points are hidden. You might be surprise to
find out that CMU uses mixture of ISP (Cogent, XO, and KINBER). These
are not normal IPS.  More surprisingly is that a working traceroute will
often show you that signal between your home and CMU goes through
Virginia (NSA). Yap that is right. CMU also uses Managed DNS
authoritative servers which they outsourced few years ago to a company
(I forgot the name but it is one of those companies managing DNS for
Pentagon and alike). At some point I remember finding one of their DNS
servers located in New York misconfigured (of course they denied that).

Anyhow in my experience your problem will eventually magically disappear
and they have nothing to do with FIOS.

Oh and yes CMU does have a black list of IP addresses that they are
blocking. They actually block entire blocks of IP addresses. In your
case the reason for loss of connection could have been simply the fact
that you got a new IP address (dhcp lease) on Monday from your ISP. That
IP address could have been from a block of IP addressed which is
currently being blocked by CMU guys for whatever reason (used recently
for example for DoS attack by an adversarial foreign nation state). The
blocking is typically temporary as those addresses are assigned to U.S.
consumers but might have been temporary high-jacked for an attack. No
Russians or Serbs for that matter don't use their IP addresses for
attach on US just like US agencies do not US addresses to attack Iran
for example. The first step is always taking control of large number of
personal computers from all over the world from incompetent Internet
Service Providers and their even more incompetent users and then staging
massive dynamical attack where machines who are attacking you appear
from nowhere and everywhere. 

Sorry I could not be of more help but I hope you had fun reading this.

Predrag

From jim at xuth.net  Sat Jul 27 14:47:02 2019
From: jim at xuth.net (Jim Leonard)
Date: Sat, 27 Jul 2019 19:47:02 +0100
Subject: Weird networking problem on home net access to CMU
In-Reply-To: <20190727174836.VkKkDCF_F%predragp@andrew.cmu.edu>
References: <CAAgoO27kAGbH+yrQWaB8KHTyiLNVpyrYr6g8sTnKB+6Js7nrpA@mail.gmail.com>
 <20190727174836.VkKkDCF_F%predragp@andrew.cmu.edu>
Message-ID: <20190727184702.GA5464@xuth.net>

I've had problems where VPNs leave my routing and dns in a bad state that won't let me access anything in clusters of addresses without going through the (now non-existent) vpn tunnel.

On Sat, Jul 27, 2019 at 01:48:36PM -0400, Predrag Punosevac wrote:
> Robert MacLachlan <robmacl at cmu.edu> wrote:
> 
> > I have verizon FIOS and having this weird problem where I can access
> > anything in the entire world except the CMU and auton nets.  Is there
> > any technical way to diagnose where the problem is, i.e. am I some how
> > blocked off of the CMU net by CMU, or is there some choke point in the
> > Verizon net, or what?
> > 
> > Does CMU have an IP address blocklist?
> > 
> > I am connecting by tethering to my phone (T-moble) which has no
> > problem, but this is of course slow.  The problem started this week,
> > was there on monday, then stopped for a while and came back maybe
> > thursday.  At first I thought that the CMU net was actually down.
> > 
> >   Rob
> 
> Hi Rob,
> 
> I have seen and experienced this first hand myself few years ago with
> Armstrong cable as my ISP. I went to a great deal of network
> troubleshooting (lot of traceroute, dig, tcpdump, taking to CMU network
> guys including off site) without getting conclusive evidence pointing to
> any particular reason.
> 
> For starters you can try to traceroute www.cmu.edu from your machine as
> well as from one of content global delivery networks. 
> 
> https://tools.keycdn.com/traceroute
> 
> You will need to use dig and whois to convert all those IP addresses to
> domain names and legal entities.  I can see that at this very moment CMU
> network is not reachable from Miami and San Francisco servers.
> Unfortunately the breaking points are hidden. You might be surprise to
> find out that CMU uses mixture of ISP (Cogent, XO, and KINBER). These
> are not normal IPS.  More surprisingly is that a working traceroute will
> often show you that signal between your home and CMU goes through
> Virginia (NSA). Yap that is right. CMU also uses Managed DNS
> authoritative servers which they outsourced few years ago to a company
> (I forgot the name but it is one of those companies managing DNS for
> Pentagon and alike). At some point I remember finding one of their DNS
> servers located in New York misconfigured (of course they denied that).
> 
> Anyhow in my experience your problem will eventually magically disappear
> and they have nothing to do with FIOS.
> 
> Oh and yes CMU does have a black list of IP addresses that they are
> blocking. They actually block entire blocks of IP addresses. In your
> case the reason for loss of connection could have been simply the fact
> that you got a new IP address (dhcp lease) on Monday from your ISP. That
> IP address could have been from a block of IP addressed which is
> currently being blocked by CMU guys for whatever reason (used recently
> for example for DoS attack by an adversarial foreign nation state). The
> blocking is typically temporary as those addresses are assigned to U.S.
> consumers but might have been temporary high-jacked for an attack. No
> Russians or Serbs for that matter don't use their IP addresses for
> attach on US just like US agencies do not US addresses to attack Iran
> for example. The first step is always taking control of large number of
> personal computers from all over the world from incompetent Internet
> Service Providers and their even more incompetent users and then staging
> massive dynamical attack where machines who are attacking you appear
> from nowhere and everywhere. 
> 
> Sorry I could not be of more help but I hope you had fun reading this.
> 
> Predrag

From predragp at andrew.cmu.edu  Sat Jul 27 15:48:27 2019
From: predragp at andrew.cmu.edu (Predrag Punosevac)
Date: Sat, 27 Jul 2019 15:48:27 -0400
Subject: Weird networking problem on home net access to CMU
In-Reply-To: <20190727184702.GA5464@xuth.net>
References: <CAAgoO27kAGbH+yrQWaB8KHTyiLNVpyrYr6g8sTnKB+6Js7nrpA@mail.gmail.com>
 <20190727174836.VkKkDCF_F%predragp@andrew.cmu.edu>
 <20190727184702.GA5464@xuth.net>
Message-ID: <20190727194827.qlKoO7DIx%predragp@andrew.cmu.edu>

Jim Leonard <jim at xuth.net> wrote:

> I've had problems where VPNs leave my routing and dns in a bad state
> that won't let me access anything in clusters of addresses without
> going through the (now non-existent) vpn tunnel. 

You are going to see a lot of that stuff once you start playing with
that mil.gov Kerberos authentication. Right Kyle :-)

Predrag



> On Sat, Jul 27, 2019 at 01:48:36PM -0400, Predrag Punosevac wrote:
> > Robert MacLachlan <robmacl at cmu.edu> wrote:
> > 
> > > I have verizon FIOS and having this weird problem where I can access
> > > anything in the entire world except the CMU and auton nets.  Is there
> > > any technical way to diagnose where the problem is, i.e. am I some how
> > > blocked off of the CMU net by CMU, or is there some choke point in the
> > > Verizon net, or what?
> > > 
> > > Does CMU have an IP address blocklist?
> > > 
> > > I am connecting by tethering to my phone (T-moble) which has no
> > > problem, but this is of course slow.  The problem started this week,
> > > was there on monday, then stopped for a while and came back maybe
> > > thursday.  At first I thought that the CMU net was actually down.
> > > 
> > >   Rob
> > 
> > Hi Rob,
> > 
> > I have seen and experienced this first hand myself few years ago with
> > Armstrong cable as my ISP. I went to a great deal of network
> > troubleshooting (lot of traceroute, dig, tcpdump, taking to CMU network
> > guys including off site) without getting conclusive evidence pointing to
> > any particular reason.
> > 
> > For starters you can try to traceroute www.cmu.edu from your machine as
> > well as from one of content global delivery networks. 
> > 
> > https://tools.keycdn.com/traceroute
> > 
> > You will need to use dig and whois to convert all those IP addresses to
> > domain names and legal entities.  I can see that at this very moment CMU
> > network is not reachable from Miami and San Francisco servers.
> > Unfortunately the breaking points are hidden. You might be surprise to
> > find out that CMU uses mixture of ISP (Cogent, XO, and KINBER). These
> > are not normal IPS.  More surprisingly is that a working traceroute will
> > often show you that signal between your home and CMU goes through
> > Virginia (NSA). Yap that is right. CMU also uses Managed DNS
> > authoritative servers which they outsourced few years ago to a company
> > (I forgot the name but it is one of those companies managing DNS for
> > Pentagon and alike). At some point I remember finding one of their DNS
> > servers located in New York misconfigured (of course they denied that).
> > 
> > Anyhow in my experience your problem will eventually magically disappear
> > and they have nothing to do with FIOS.
> > 
> > Oh and yes CMU does have a black list of IP addresses that they are
> > blocking. They actually block entire blocks of IP addresses. In your
> > case the reason for loss of connection could have been simply the fact
> > that you got a new IP address (dhcp lease) on Monday from your ISP. That
> > IP address could have been from a block of IP addressed which is
> > currently being blocked by CMU guys for whatever reason (used recently
> > for example for DoS attack by an adversarial foreign nation state). The
> > blocking is typically temporary as those addresses are assigned to U.S.
> > consumers but might have been temporary high-jacked for an attack. No
> > Russians or Serbs for that matter don't use their IP addresses for
> > attach on US just like US agencies do not US addresses to attack Iran
> > for example. The first step is always taking control of large number of
> > personal computers from all over the world from incompetent Internet
> > Service Providers and their even more incompetent users and then staging
> > massive dynamical attack where machines who are attacking you appear
> > from nowhere and everywhere. 
> > 
> > Sorry I could not be of more help but I hope you had fun reading this.
> > 
> > Predrag

From awd at cs.cmu.edu  Mon Jul 29 05:02:18 2019
From: awd at cs.cmu.edu (Artur Dubrawski)
Date: Mon, 29 Jul 2019 05:02:18 -0400
Subject: Fwd: Thesis Defense - July 30th - Eric Lei - Multi-view Relationships
 for Analytics and Inference
In-Reply-To: <ca3fb1b6-111c-6487-8d4b-ecb1493cdf7c@cmu.edu>
References: <ca3fb1b6-111c-6487-8d4b-ecb1493cdf7c@cmu.edu>
Message-ID: <CAJvAoys_g7Na8kiKixAEV=k8X+atmYV0cAdC1nmw2Tn2s4Dgag@mail.gmail.com>

Dear Autonians,

Join and see Eric become a Doctor!

Cheers,
Artur



---------- Forwarded message ---------
From: Diane Stidle <stidle at andrew.cmu.edu>
Date: Mon, Jul 22, 2019 at 1:35 PM
Subject: Thesis Defense - July 30th - Eric Lei - Multi-view Relationships
for Analytics and Inference
To: ml-seminar at cs.cmu.edu <ML-SEMINAR at cs.cmu.edu>, Mario Berges <
marioberges at cmu.edu>, <labov1 at llnl.gov>


*Thesis Defense*

Date: July 30, 2019
Time: 3:00pm
Place: GHC 8102
PhD Candidate: Eric Lei

*Title: **Multi-view Relationships for Analytics and Inference*

Abstract:

An interesting area of machine learning is methods for multi-view data,
relational data whose features have been partitioned. Multi-view learning
exploits relationships between views, giving it certain advantages over
traditional single-view techniques, which may struggle to find these
relationships or only learn them implicitly. These relationships are often
especially salient in understanding the data or performing prediction. This
work explores an underutilized approach in multi-view learning: to focus on
multi-view relationships---the variables that govern relations between
views---themselves as units of analysis.  We investigate how this approach
impacts analytics and inference in ways that standard multi-view and
single-view learning cannot. We hypothesize that by ignoring relations
between views or factoring them in only indirectly, standard approaches
risk overlooking key structure. Accordingly, our goal is to investigate the
extent multi-view relationships can be characterized and employed as units
of analysis in descriptive analytics and inference. We present novel
methods to do so, either using domain knowledge or by learning from data,
which reveal structure that alternative methods do not or have competitive
performance with the state of the art. Empirical results are presented in
several application domains. First, we use domain knowledge to assume a
known form for multi-view relationships in the task of gamma source
detection. We aggregate the views by filtering their inferences
collectively to perform classification. Second, we assume multi-view
relationships are linear and learn them from data in a different approach
toward gamma source detection. Our method detects anomalies when these
relationships are disrupted. Third, we relax the assumption of linearity
and propose a novel clustering method that finds cluster-wise linear
relationships. This method discovers explanatory structure in a medical
problem. Fourth, we extend this method to classification and demonstrate
its superior performance on a load monitoring problem.

*Thesis Committee: *
Artur Dubrawski (Chair)
Barnabas Poczos
Mario Berges
Simon Labov (Lawrence Livermore National Laboratory)

-- 
Diane Stidle
Graduate Programs Manager
Machine Learning Department
Carnegie Mellon Universitystidle at cmu.edu
412-268-1299
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190729/fd2da054/attachment.html>

From boecking at andrew.cmu.edu  Mon Jul 29 16:04:30 2019
From: boecking at andrew.cmu.edu (Benedikt Boecking)
Date: Mon, 29 Jul 2019 16:04:30 -0400
Subject: MSCOCO
Message-ID: <EB6032BC-ABB1-4214-830F-7F3CEAC3481E@andrew.cmu.edu>

Hi all,

Does anyone at the Auton Lab have MSCOCO on one of the (gpu) compute nodes? If so, would you be willing to give me read access? I want to avoid unnecessarily bloating the servers with duplicates. 

Best,
Ben




From awd at cs.cmu.edu  Tue Jul 30 09:29:31 2019
From: awd at cs.cmu.edu (Artur Dubrawski)
Date: Tue, 30 Jul 2019 09:29:31 -0400
Subject: Fwd: Reminder - Thesis Defense - July 30th - Eric Lei - Multi-view
 Relationships for Analytics and Inference
In-Reply-To: <9aedc3a0-d450-3bd2-e35a-a1769cef5697@cmu.edu>
References: <9aedc3a0-d450-3bd2-e35a-a1769cef5697@cmu.edu>
Message-ID: <CAJvAoyv3Oqo8w9xErtceRcMuEOnCSEU856GKhcGy_hTktWybTA@mail.gmail.com>

Just a reminder, see you there!

Artur

---------- Forwarded message ---------
From: Diane Stidle <stidle at andrew.cmu.edu>
Date: Tue, Jul 30, 2019, 9:28 AM
Subject: Reminder - Thesis Defense - July 30th - Eric Lei - Multi-view
Relationships for Analytics and Inference
To: ml-seminar at cs.cmu.edu <ML-SEMINAR at cs.cmu.edu>, Mario Berges <
marioberges at cmu.edu>, <labov1 at llnl.gov>


*Thesis Defense*

Date: July 30, 2019
Time: 3:00pm
Place: GHC 8102
PhD Candidate: Eric Lei

*Title: **Multi-view Relationships for Analytics and Inference*

Abstract:

An interesting area of machine learning is methods for multi-view data,
relational data whose features have been partitioned. Multi-view learning
exploits relationships between views, giving it certain advantages over
traditional single-view techniques, which may struggle to find these
relationships or only learn them implicitly. These relationships are often
especially salient in understanding the data or performing prediction. This
work explores an underutilized approach in multi-view learning: to focus on
multi-view relationships---the variables that govern relations between
views---themselves as units of analysis.  We investigate how this approach
impacts analytics and inference in ways that standard multi-view and
single-view learning cannot. We hypothesize that by ignoring relations
between views or factoring them in only indirectly, standard approaches
risk overlooking key structure. Accordingly, our goal is to investigate the
extent multi-view relationships can be characterized and employed as units
of analysis in descriptive analytics and inference. We present novel
methods to do so, either using domain knowledge or by learning from data,
which reveal structure that alternative methods do not or have competitive
performance with the state of the art. Empirical results are presented in
several application domains. First, we use domain knowledge to assume a
known form for multi-view relationships in the task of gamma source
detection. We aggregate the views by filtering their inferences
collectively to perform classification. Second, we assume multi-view
relationships are linear and learn them from data in a different approach
toward gamma source detection. Our method detects anomalies when these
relationships are disrupted. Third, we relax the assumption of linearity
and propose a novel clustering method that finds cluster-wise linear
relationships. This method discovers explanatory structure in a medical
problem. Fourth, we extend this method to classification and demonstrate
its superior performance on a load monitoring problem.

*Thesis Committee: *
Artur Dubrawski (Chair)
Barnabas Poczos
Mario Berges
Simon Labov (Lawrence Livermore National Laboratory)

-- 
Diane Stidle
Graduate Programs Manager
Machine Learning Department
Carnegie Mellon Universitystidle at cmu.edu
412-268-1299
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190730/9bdebcbd/attachment.html>

From d.c.howarth at gmail.com  Tue Jul 30 15:09:02 2019
From: d.c.howarth at gmail.com (Dan Howarth)
Date: Tue, 30 Jul 2019 15:09:02 -0400
Subject: lov5 scratch 99% full
Message-ID: <CACG4Uo-gJO96jzvKWVj9JedoHNu9tc6sWp-Qo9A13awhc1Q0ZA@mail.gmail.com>

Hello all,

lov5 scratch is 99% full, please take a moment to see if you can remove
anything

Thanks!
Dan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/autonlab-users/attachments/20190730/bdf609bb/attachment.html>