From ashwinpp at stanford.edu Sun Apr 2 02:07:35 2017
From: ashwinpp at stanford.edu (Ashwin Pradeep Paranjape)
Date: Sun, 2 Apr 2017 09:07:35 +0000
Subject: [theory-seminar] Sanjeev Arora talks about GANs
Message-ID:
5th April, Wednesday 3-4 pm
Gates 219 (2A open area)
Topic -
Generalization and Equilibrium in Generative Adversarial Nets (GANs)
Abstract-
This paper makes progress on several open theoretical issues related to Generative Adversarial Networks. A definition is provided for what it means for the training to?generalize, and it is shown that generalization is not guaranteed for the popular distances between distributions such as Jensen-Shannon or?Wasserstein. We introduce a new metric called?neural net distance?for which generalization does occur. We also show that an approximate pure equilibrium in the 2- player game exists for a natural training objective (Wasserstein). Showing such a result has been an open problem (for any training objective).
Finally, the above theoretical ideas lead us to pro- pose a new training protocol,?MIX+GAN, which can be combined with any existing method. We present experiments showing that it stabilizes and improves some existing methods.
Joint work with Rong Ge, Yingyu Liang, Tengyu Ma, Yi Zhang.?
Speaker Bio -
Sanjeev Arora is Charles C. Fitzmorris Professor of Computer Science at Princeton University. He got his PhD at UC Berkeley in 1994. His research area spans several areas of theoretical Computer Science, including computational complexity and algorithm design, and theoretical problems in machine learning. He has received the ACM Doctoral Dissertation Award (1995), ACM-EATCS G?del Prize (in 2001 and 2010), Packard Fellowship (1997), the ACM Infosys Foundation Award in the Computing Sciences (2012), the Fulkerson Prize (2012), and the Simons Investigator Award (2012).
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/calendar
Size: 3561 bytes
Desc: not available
URL:
From limingchao333 at gmail.com Mon Apr 3 11:35:18 2017
From: limingchao333 at gmail.com (mingchao li)
Date: Mon, 3 Apr 2017 11:35:18 -0700
Subject: [theory-seminar] Cooperation from a veture capital
Message-ID:
Dear Professor,
I?m a venture capitalist from China. Our Venture Capital ?Amphora
Investment is one of the top 20s in China. Now we are forcing on Artificial
Intelligence and Precision medical care.
Afanti(????which is a student-oriented intelligent learning mobile phone
applications for users in the learning process encountered problems
provides the appropriate solution. At the end of February 2016, the number
of users registered is more than 29.08 million people, 1.24 million daily
active users. Tencent invested the C round, the current valuation is 3
billion RMB. We invested it in the early 2014 in the Angel Round, It is one
of our portfolio in AI.
In view of your extraordinary academic achievements, whether you have
some entrepreneurial projects seeking for funds, or you are interested in
being a technical adviser in our Venture agency, or some other
opportunities you think we can cooperate,please contact me when you are
convenient. I?m very interested in talking with you face-to ?face or on the
phone.
I?m looking forward to your replying.
William Li Amphora Investment Director
US office: 1155 Tasman dr.Sunnyvale.CA.94089.USA
Mobile:1+6269997602
Wechat?1184628158
Official Website of China?www.vcpe.hk
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From gvaliant at cs.stanford.edu Tue Apr 4 11:14:29 2017
From: gvaliant at cs.stanford.edu (Gregory Valiant)
Date: Tue, 4 Apr 2017 11:14:29 -0700
Subject: [theory-seminar] no thursday seminar--post-deadline party instead
Message-ID:
Hi Friends,
Because of the Thursday, 5pm FOCS deadline, we won't be having a
theory seminar this week. Instead, we'll have our usual post-deadline
celebration.
5pm, Gates 463a. I'll bring the drinks and snacks.
Hope to see you all there,
-g
From michael.kim at cs.stanford.edu Wed Apr 5 10:20:17 2017
From: michael.kim at cs.stanford.edu (Michael Kim)
Date: Wed, 5 Apr 2017 10:20:17 -0700
Subject: [theory-seminar] Theory Lunch 4/6 - Organizational Meeting
In-Reply-To:
References:
Message-ID:
Hi all,
Good luck to those finishing up submissions! We will have theory lunch
tomorrow in Gates 463 at noon as usual. There will not be a talk tomorrow,
but instead we'll have sign-ups for the rest of the quarter as well as any
announcements.
If you can't make tomorrow's lunch but would like to talk this quarter,
just shoot me an email with your availabilities!
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From ashwinpp at stanford.edu Wed Apr 5 15:00:08 2017
From: ashwinpp at stanford.edu (Ashwin Pradeep Paranjape)
Date: Wed, 5 Apr 2017 22:00:08 +0000
Subject: [theory-seminar] Fw: Sanjeev Arora talks about GANs
In-Reply-To:
References:
Message-ID:
Happening now
________________________________
From: Ashwin Pradeep Paranjape
Sent: Sunday, April 2, 2017 2:07 AM
To: ai-all at cs.stanford.edu; theory-seminar at lists.stanford.edu; CS Seminars; Christopher Manning
Subject: [theory-seminar] Sanjeev Arora talks about GANs
When: Wednesday, April 5, 2017 6:00 AM-7:00 AM.
Where: Gates 213
5th April, Wednesday 3-4 pm
Gates 219 (2A open area)
Topic -
Generalization and Equilibrium in Generative Adversarial Nets (GANs)
Abstract-
This paper makes progress on several open theoretical issues related to Generative Adversarial Networks. A definition is provided for what it means for the training to generalize, and it is shown that generalization is not guaranteed for the popular distances between distributions such as Jensen-Shannon or Wasserstein. We introduce a new metric called neural net distance for which generalization does occur. We also show that an approximate pure equilibrium in the 2- player game exists for a natural training objective (Wasserstein). Showing such a result has been an open problem (for any training objective).
Finally, the above theoretical ideas lead us to pro- pose a new training protocol, MIX+GAN, which can be combined with any existing method. We present experiments showing that it stabilizes and improves some existing methods.
Joint work with Rong Ge, Yingyu Liang, Tengyu Ma, Yi Zhang.
Speaker Bio -
Sanjeev Arora is Charles C. Fitzmorris Professor of Computer Science at Princeton University. He got his PhD at UC Berkeley in 1994. His research area spans several areas of theoretical Computer Science, including computational complexity and algorithm design, and theoretical problems in machine learning. He has received the ACM Doctoral Dissertation Award (1995), ACM-EATCS G?del Prize (in 2001 and 2010), Packard Fellowship (1997), the ACM Infosys Foundation Award in the Computing Sciences (2012), the Fulkerson Prize (2012), and the Simons Investigator Award (2012).
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/calendar
Size: 3790 bytes
Desc: not available
URL:
From gvaliant at cs.stanford.edu Thu Apr 6 12:04:06 2017
From: gvaliant at cs.stanford.edu (Gregory Valiant)
Date: Thu, 6 Apr 2017 12:04:06 -0700
Subject: [theory-seminar] post-deadline party @ 5pm today
Message-ID:
Hi Friends,
We'll be having our usual post-deadline festivities today at 5pm in
Gates 463a. Please stop by for snacks and libations.
Cheers,
-g
From michael.kim at cs.stanford.edu Thu Apr 6 13:04:13 2017
From: michael.kim at cs.stanford.edu (Michael Kim)
Date: Thu, 6 Apr 2017 13:04:13 -0700
Subject: [theory-seminar] Food in 463A
Message-ID:
There is plenty of Ike's Sandwiches leftover in Gates 463. I've booked the
room for the remainder of the afternoon, so that if you're working towards
the deadline but need some nourishment, come on by to eat and work!
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From michael.kim at cs.stanford.edu Wed Apr 12 10:05:53 2017
From: michael.kim at cs.stanford.edu (Michael Kim)
Date: Wed, 12 Apr 2017 10:05:53 -0700
Subject: [theory-seminar] Theory Lunch 4/13 - Don Knuth
Message-ID:
Hi all -
Tomorrow, we will have theory lunch as usual at noon in Gates 463. Don
will be our first speaker of the quarter! He plans on telling us about a
new, "motley" notion that should be of interest to many.
Hope to see you there!
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From kentlee at stanford.edu Thu Apr 13 14:10:00 2017
From: kentlee at stanford.edu (Kent Lee)
Date: Thu, 13 Apr 2017 21:10:00 +0000
Subject: [theory-seminar] Where were burritos from for todays seminar?
Message-ID:
They were good!
From ilanama at stanford.edu Mon Apr 17 14:29:41 2017
From: ilanama at stanford.edu (Ilana Arbisser)
Date: Mon, 17 Apr 2017 14:29:41 -0700
Subject: [theory-seminar] Fwd: [stanford_complexity_group] Spring Seminar:
John Baez, Biology as Information Dynamics, April 20th
In-Reply-To:
References:
Message-ID:
John Baez, Professor of Mathematics at UC Riverside will be giving the
Stanford Complexity Group Spring Seminar:
*Biology as Information Dynamics*
If biology is the study of self-replicating entities, and we want to
understand the role of information, it makes sense to see how information
theory is connected to the 'replicator equation' ? a simple model of
population dynamics for self-replicating entities. The relevant concept of
information turns out to be the information of one probability distribution
relative to another, also known as the Kullback?Liebler divergence. Using
this we can get a new outlook on free energy, see evolution as a learning
process, and give a clearer, more general formulation of Fisher's
fundamental theorem of natural selection.
The seminar will be April 20th, at 4:20PM in Clark S361.
Please forward to any lists that would find this seminar relevant.
If you are interested in meeting with Prof. Baez earlier in the day, please
let me know.
Ilana Arbisser
PhD Candidate
Evolutionary Biology
_______________________________________________
stanford_complexity_group mailing list
stanford_complexity_group at lists.stanford.edu
https://mailman.stanford.edu/mailman/listinfo/stanford_complexity_group
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ComplexitySpring2017.png
Type: image/png
Size: 115533 bytes
Desc: not available
URL:
From ashwinpp at stanford.edu Tue Apr 18 14:59:11 2017
From: ashwinpp at stanford.edu (Ashwin Pradeep Paranjape)
Date: Tue, 18 Apr 2017 21:59:11 +0000
Subject: [theory-seminar] Algorithms reading group
Message-ID:
Hi all,
We shall be starting our algorithms reading group (a.k.a Algorithms, Geometry and Learning) with an organizational meeting. I've set up a doodle poll to figure out a meeting time for just this week.
http://doodle.com/poll/gxadybpz3pw754cq#table
I have also set up a doodle poll for the recurrent reading group meetings for coming weeks (excluding this week). It shows dates for next week but it really represents all the coming weeks.
http://doodle.com/poll/t224hg3ba9srdkvd
This week we shall be discussing which papers to read this quarter, in continuation to a few STOC papers that were discussed last quarter. We shall also be deciding the on the presenter and date of presentation. Hope to see you there!
If you have any interesting papers in mind, please put them in the following google spreadsheet
https://docs.google.com/spreadsheets/d/1_59eltRyMBvyHNEMAMwZA-g0VqNvre6YTPaz5wUtRv8/edit?usp=sharing
Thanks
Ashwin
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From michael.kim at cs.stanford.edu Wed Apr 19 09:52:04 2017
From: michael.kim at cs.stanford.edu (Michael Kim)
Date: Wed, 19 Apr 2017 09:52:04 -0700
Subject: [theory-seminar] Theory Lunch 4/20 -- Speed Talks!
Message-ID:
Hi all,
Please join us tomorrow for theory lunch -- same time, same place. As a
reminder, this week, all attendees should prepare a 5 minute speed talk
that will be part of the theory lunch program. It would be good to present
something you've been working on or thinking about lately that the group
might be interested in.
Hope to see you there!
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From ilanama at stanford.edu Thu Apr 20 11:22:52 2017
From: ilanama at stanford.edu (Ilana Arbisser)
Date: Thu, 20 Apr 2017 11:22:52 -0700
Subject: [theory-seminar] Fwd: [stanford_complexity_group] Spring Seminar:
John Baez, Biology as Information Dynamics, April 20th
In-Reply-To:
References:
Message-ID:
Reminder: this event is happening *today*! If you would like to meet with
Professor Baez on Friday, please email aaronjg at stanford.edu.
Thanks,
Ilana
---------- Forwarded message ----------
From: Aaron Goodman
Date: Mon, Apr 3, 2017 at 3:45 PM
Subject: [stanford_complexity_group] Spring Seminar: John Baez, Biology as
Information Dynamics, April 20th
To: stanford_complexity_group at lists.stanford.edu
John Baez, Professor of Mathematics at UC Riverside will be giving the
Stanford Complexity Group Spring Seminar:
*Biology as Information Dynamics*
If biology is the study of self-replicating entities, and we want to
understand the role of information, it makes sense to see how information
theory is connected to the 'replicator equation' ? a simple model of
population dynamics for self-replicating entities. The relevant concept of
information turns out to be the information of one probability distribution
relative to another, also known as the Kullback?Liebler divergence. Using
this we can get a new outlook on free energy, see evolution as a learning
process, and give a clearer, more general formulation of Fisher's
fundamental theorem of natural selection.
The seminar will be April 20th, at 4:20PM in Clark S361.
Please forward to any lists that would find this seminar relevant.
If you are interested in meeting with Prof. Baez earlier in the day, please
let me know.
Aaron Goodman
PhD Candidate
Evolutionary Biology
_______________________________________________
stanford_complexity_group mailing list
stanford_complexity_group at lists.stanford.edu
https://mailman.stanford.edu/mailman/listinfo/stanford_complexity_group
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ComplexitySpring2017.png
Type: image/png
Size: 115533 bytes
Desc: not available
URL:
From ashwinpp at stanford.edu Thu Apr 20 11:58:17 2017
From: ashwinpp at stanford.edu (Ashwin Pradeep Paranjape)
Date: Thu, 20 Apr 2017 18:58:17 +0000
Subject: [theory-seminar] [algorithms-geometry-learning] Algorithms
reading group
In-Reply-To:
References:
Message-ID:
We'll be having our organisational meeting tomorrow (Friday 21st April) at 11 pm. In case you haven't filled your preferred dates for the rest of the quarter, please do so at
http://doodle.com/poll/t224hg3ba9srdkvd
See you tomorrow,
Ashwin
From: Ashwin Pradeep Paranjape
Sent: Tuesday, April 18, 2:59 PM
Subject: [algorithms-geometry-learning] Algorithms reading group
To: algorithms_reading_group at lists.stanford.edu, algorithms-geometry-learning at lists.stanford.edu, theory-students at cs.stanford.edu, theory-seminar at lists.stanford.edu
Hi all,
We shall be starting our algorithms reading group (a.k.a Algorithms, Geometry and Learning) with an organizational meeting. I've set up a doodle poll to figure out a meeting time for just this week.
http://doodle.com/poll/gxadybpz3pw754cq#table
I have also set up a doodle poll for the recurrent reading group meetings for coming weeks (excluding this week). It shows dates for next week but it really represents all the coming weeks.
http://doodle.com/poll/t224hg3ba9srdkvd
This week we shall be discussing which papers to read this quarter, in continuation to a few STOC papers that were discussed last quarter. We shall also be deciding the on the presenter and date of presentation. Hope to see you there!
If you have any interesting papers in mind, please put them in the following google spreadsheet
https://docs.google.com/spreadsheets/d/1_59eltRyMBvyHNEMAMwZA-g0VqNvre6YTPaz5wUtRv8/edit?usp=sharing
Thanks
Ashwin
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From ashwinpp at stanford.edu Thu Apr 20 12:05:12 2017
From: ashwinpp at stanford.edu (Ashwin Pradeep Paranjape)
Date: Thu, 20 Apr 2017 19:05:12 +0000
Subject: [theory-seminar] [algorithms-geometry-learning] Algorithms
reading group
In-Reply-To:
References: ,
Message-ID:
Oops, I meant 11 am. Sorry for the confusion.
________________________________
From: Ashwin Pradeep Paranjape
Sent: Thursday, April 20, 2017 11:58:17 AM
To: algorithms_reading_group at lists.stanford.edu; algorithms-geometry-learning at lists.stanford.edu; theory-students at cs.stanford.edu; theory-seminar at lists.stanford.edu
Subject: Re: [algorithms-geometry-learning] Algorithms reading group
We'll be having our organisational meeting tomorrow (Friday 21st April) at 11 pm. In case you haven't filled your preferred dates for the rest of the quarter, please do so at
http://doodle.com/poll/t224hg3ba9srdkvd
See you tomorrow,
Ashwin
From: Ashwin Pradeep Paranjape
Sent: Tuesday, April 18, 2:59 PM
Subject: [algorithms-geometry-learning] Algorithms reading group
To: algorithms_reading_group at lists.stanford.edu, algorithms-geometry-learning at lists.stanford.edu, theory-students at cs.stanford.edu, theory-seminar at lists.stanford.edu
Hi all,
We shall be starting our algorithms reading group (a.k.a Algorithms, Geometry and Learning) with an organizational meeting. I've set up a doodle poll to figure out a meeting time for just this week.
http://doodle.com/poll/gxadybpz3pw754cq#table
I have also set up a doodle poll for the recurrent reading group meetings for coming weeks (excluding this week). It shows dates for next week but it really represents all the coming weeks.
http://doodle.com/poll/t224hg3ba9srdkvd
This week we shall be discussing which papers to read this quarter, in continuation to a few STOC papers that were discussed last quarter. We shall also be deciding the on the presenter and date of presentation. Hope to see you there!
If you have any interesting papers in mind, please put them in the following google spreadsheet
https://docs.google.com/spreadsheets/d/1_59eltRyMBvyHNEMAMwZA-g0VqNvre6YTPaz5wUtRv8/edit?usp=sharing
Thanks
Ashwin
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From moses at cs.stanford.edu Mon Apr 24 23:10:24 2017
From: moses at cs.stanford.edu (Moses Charikar)
Date: Mon, 24 Apr 2017 23:10:24 -0700
Subject: [theory-seminar] Afonso Bandeira, tomorrow (Apr 25), 4:30pm
Message-ID:
Some of you may be interested in the Stats seminar talk tomorrow (Tue):
--Moses
Tuesday, April 25, 4:30pm: Statistics Seminar, Sloan Mathematics Center
Room 380X
Afonso Bandeira, Courant Institute, NYU
"On phase transitions for spiked random matrix and tensor models"
Abstract:
A central problem of random matrix theory is to understand the eigenvalues
of spiked random matrix models, in which a prominent eigenvector (or low
rank structure) is planted into a random matrix. These distributions form
natural statistical models for principal component analysis (PCA) problems
throughout the sciences, where the goal is often to recover or detect the
planted low rank structured. In this talk we discuss fundamental
limitations of statistical methods to perform these tasks and methods that
outperform PCA at it. Emphasis will be given to low rank structures arising
in synchronization problems. Time permitting, analogous results for spiked
tensor models will also be discussed.
This talk is based on joint work with Amelia Perry, Alex Wein, and Ankur
Moitra.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From michael.kim at cs.stanford.edu Wed Apr 26 12:04:15 2017
From: michael.kim at cs.stanford.edu (Michael Kim)
Date: Wed, 26 Apr 2017 12:04:15 -0700
Subject: [theory-seminar] Theory Lunch 4/27 -- Huacheng Yu
Message-ID:
Hi all,
There will be theory lunch tomorrow at noon in 463 as usual. Huacheng will
be our speaker and will tell us about some exciting results on cell-probe
lower bounds. (Shameless plug for theory extra-curricular activities:
Huacheng et al. started collaborating on this problem during the Stanford
theory retreat!)
Look forward to seeing you there!
******
Speaker: Huacheng Yu
Title: Cell-Probe Lower Bounds from Online Communication Complexity
Abstract: In this work, we introduce an online model for communication
complexity. Analogous to how online algorithms receive their input piece-by-
piece, our model presents one of the players Bob his input piece-by-piece,
and
has the players Alice and Bob cooperate to compute a result it presents Bob
with the next piece. This model has a closer and more natural correspondence
to dynamic data structures than the classic communication models do and
hence
presents a new perspective on data structures.
We apply the online communication model to data structure lower bounds by
studying the Group Range Problem, a dynamic data structure problem. This
problem admits an O(log n)-time worst-case data structure. Using online
communication complexity, we prove a tight cell-probe lower bound: spending
o(log n) (even amortized) time per operation results in at best an
exp(-delta^2 n) probability of correctly answering a (1/2+delta)-fraction of
the n queries.
In this talk, I will present a lower bound for the online set intersection
problem in the online communication model, demonstrating a general approach
for proving online communication lower bounds. The online communication
model
prevents a batching trick that classic communication complexity allows, and
yields a higher lower bound.
(The online set intersection problem is posted as an open problem in the
2015
theory retreat.)
Joint work with Josh Alman and Joshua R. Wang.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From moses at cs.stanford.edu Sun Apr 30 16:42:13 2017
From: moses at cs.stanford.edu (Moses Charikar)
Date: Sun, 30 Apr 2017 16:42:13 -0700
Subject: [theory-seminar] short talks at TOCA-SV, May 12
Message-ID:
Theory folks,
As mentioned before, we have our 2nd TOCA-SV meeting on Friday May 12.
Google is hosting the event. Registration and other details will be sent
out shortly. Hoping that there will be a strong Stanford representation. We
will coordinate travel from campus to Google and back.
I'd like to get a list of speakers for a short talk session. Speakers will
have 5 min. Just enough time to communicate a cool result, a cool theorem,
sing a song perhaps? ... a great opportunity to introduce yourself to the
Bay Area theory community.
Sign up here:
(Just names and email addresses; title if you have it -- we won't hold you
to it. If we are oversubscribed, we may have to limit the speaker list to
the earliest signups)
https://docs.google.com/document/d/1HBAuV_Zxf8k6XrR1-6czxKdrvupK-LKSnONrJDpIrgM/edit?usp=sharing
Cheers,
Moses
-------------- next part --------------
An HTML attachment was scrubbed...
URL: