From ccanonne at cs.stanford.edu Wed Sep 4 09:59:01 2019
From: ccanonne at cs.stanford.edu (=?UTF-8?Q?Cl=c3=a9ment_Canonne?=)
Date: Wed, 4 Sep 2019 09:59:01 -0700
Subject: [theory-seminar] Fwd: Thursday,
September 5 -- Adi Shamir: A Simple Explanation for the Mysterious
Existence of Adversarial Examples with Small Hamming Distance
In-Reply-To:
References:
Message-ID: <4035d541-7bde-74e2-22bd-48afb932d315@cs.stanford.edu>
Hi,
In case you didn't see that on another ailing-list: Adi Shamir will be
speaking tomorrow in the security seminar. (Title and abstract below.)
Best,
-- Cl?ment
-------- Forwarded Message --------
Subject: Fwd: Thursday, September 5 -- Adi Shamir: A Simple Explanation
for the Mysterious Existence of Adversarial Examples with Small Hamming
Distance
Date: Wed, 4 Sep 2019 09:56:09 -0700
From: Saba Eskandarian
To: security-seminar at lists.stanford.edu
Reminder: we're fortunate to have Turing award winner Adi Shamir
speaking at the security seminar tomorrow!
---------- Forwarded message ---------
From: *Saba Eskandarian* >
Date: Sat, Aug 31, 2019 at 11:13 PM
Subject: Thursday, September 5 -- Adi Shamir: A Simple Explanation for
the Mysterious Existence of Adversarial Examples with Small Hamming Distance
To: >
? ? A Simple Explanation for the Mysterious Existence of
? ? ? Adversarial Examples with Small Hamming Distance
? ? ? ? ? ? ? ? ? ? ? ? ?Adi Shamir
? ? ? ? ? ? ? ? Thursday, September 5, 2019
? ? ? ? ? ? ? ? ? ? ? ?Talk at 4:15pm
? ? ? ? ? ? ? ? ? ? ? ? ?Gates 104
Abstract:
The existence of adversarial examples in which tiny changes in the
input can fool well trained neural networks has many applications
and implications in object recognition, autonomous driving, cyber
security, etc.
However, it is still far from being understood why such examples exist,
and which parameters determine the number of input coordinates one
has to change in order to mislead the network.
In this talk I will describe a simple mathematical framework which
enables us to think about this problem from a fresh perspective,
turning the existence of adversarial examples from a baffling
phenomenon into a natural consequence of the geometry of R^n with
the L_0 (Hamming) metric, which can be quantitatively analyzed.
From yuvalwig at stanford.edu Mon Sep 16 15:19:32 2019
From: yuvalwig at stanford.edu (Yuval Wigderson)
Date: Mon, 16 Sep 2019 15:19:32 -0700
Subject: [theory-seminar] Fwd: Fall DRP Call for Mentors
In-Reply-To:
References:
Message-ID:
Dear theory grad students,
The math department runs a "directed reading program", where undergrads
meet one-on-one with grad student mentors for an informal reading course.
We usually get several undergrad applicants who are interested in CS-y
topics, and we often can't match them with a mentor because not enough math
grad students know the relevant topics well enough. So if you're interested
in being a mentor, we'd love to have you! Details are below.
Best,
Yuval
---------- Forwarded message ---------
From: Vivian Zieve Kuperberg
Date: Mon, Sep 16, 2019 at 2:56 PM
Subject: Fall DRP Call for Mentors
To: mathstudents at lists.stanford.edu
Dear friends,
tl;dr: if you want to DRP winter quarter, please fill out the following
Google form by Thursday, September 26th at 5pm:
https://forms.gle/3P2hbtmb7aS8QnmS9
Some graduate students in our department organize a "Directed Reading
Program," aimed at connecting undergraduate students with graduate student
"mentors" to informally learn some mathematics over the course of a quarter.
We are gearing up for the next quarter of the program and would like to
know who is interested in mentoring. The time commitment is about one hour
per week meeting with an undergraduate student, plus attending an
end-of-quarter symposium, either during tenth week of fall quarter or first
week of winter quarter. (The mentoring commitment is just one quarter at a
time.)
*If you want to mentor this quarter, please fill out the following Google
form: https://forms.gle/3P2hbtmb7aS8QnmS9
. We want your response by Thursday,
September 26th, at 5pm.*
If you have no idea what I'm talking about and want to know what is going
on, please see our fancy website at mathdrp.stanford.edu. The current
organizers are Alexander Dunlap, Cole Graham, Vivian Kuperberg, and Yuval
Wigderson; you can email or approach any of us with questions (please do!).
Given our current funding sources, mentoring will continue to be on an
unpaid basis for the present. We are pursuing opportunities to provide
mentors with modest stipends. We have funding for mentors to take their
students out to lunch once during the quarter (i.e. *you get a free
lunch!!!!*) and also to buy each student a relevant book to read.
Best,
Vivian
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From aviad at cs.stanford.edu Mon Sep 16 23:08:56 2019
From: aviad at cs.stanford.edu (Aviad Rubinstein)
Date: Mon, 16 Sep 2019 23:08:56 -0700
Subject: [theory-seminar] Wanted: theory seminar student co-organizer
Message-ID:
Hi Theory students,
Li-Yang and I will be faculty co-organizers of the theory seminar this
year, and we're looking for one or two outstanding and reliable student
co-organizers to do all the hard work.
Being a student co-organizer is a great way to interact with our celebrity
speakers and influence the choice of snacks.
This is also a great opportunity to thank Mary and Ofir, the outgoing
organizers, for their exemplary service. Thank you!
Cheers,
Aviad
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From nehgupta at stanford.edu Wed Sep 18 11:23:54 2019
From: nehgupta at stanford.edu (Neha Gupta)
Date: Wed, 18 Sep 2019 18:23:54 +0000
Subject: [theory-seminar] Theory Lunch - Talk Sign Ups
Message-ID:
Hi everyone,
The theory lunch for the fall quarter would begin on 26th September.
If you would like to give a talk, please sign up in this document -
https://docs.google.com/document/d/1S0QcDMTn-JRaP1cRFRihZyeNfHUmpYLkyORKMcGIBgY/edit .
Thanks,
Neha
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From nehgupta at stanford.edu Wed Sep 18 11:23:54 2019
From: nehgupta at stanford.edu (Neha Gupta)
Date: Wed, 18 Sep 2019 18:23:54 +0000
Subject: [theory-seminar] Theory Lunch - Talk Sign Ups
Message-ID:
Hi everyone,
The theory lunch for the fall quarter would begin on 26th September.
If you would like to give a talk, please sign up in this document -
https://docs.google.com/document/d/1S0QcDMTn-JRaP1cRFRihZyeNfHUmpYLkyORKMcGIBgY/edit .
Thanks,
Neha
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From ccanonne at cs.stanford.edu Thu Sep 19 08:39:11 2019
From: ccanonne at cs.stanford.edu (=?UTF-8?Q?Cl=c3=a9ment_Canonne?=)
Date: Thu, 19 Sep 2019 08:39:11 -0700
Subject: [theory-seminar] TCS+ talk: Wednesday, September 25, Mark Sellke,
Stanford
Message-ID: <9d41051c-6ae1-4eec-fbd0-b47377552170@cs.stanford.edu>
Hi all,
There will be a TCS+ -- the first of the Fall season -- next Wednesday,
with Mark Sellke telling us (online, but not from very far) about the
recent developments in Chasing Convex Bodies. (abstract below)
It'll be Wednesday, September 25th at 10:00 AM in Gates 463A. Breakfast
at 9:55, talk at 10 on the wall!
Best,
-- Cl?ment
-------------------------------
Speaker: Mark Sellke (Stanford)
Title: Chasing Convex Bodies
Abstract: I will explain our recent understanding of the chasing convex
bodies problem posed by Friedman and Linial in 1991. In this problem, an
online player receives a request sequence K_1,...,K_T of convex sets in
d dimensional space and moves his position online into each requested
set. The player's movement cost is the length of the resulting path.
Chasing convex bodies asks if there an online algorithm with cost
competitive against the offline optimal path. This is both an
challenging metrical task system and (equivalent to) a competitive
analysis view on online convex optimization.
This problem was open for d>2 until last year but has recently been
solved twice. The first solution gives a 2^{O(d)} competitive algorithm
while the second gives a nearly optimal min(d,sqrt(d*log(T)))
competitive algorithm for T requests. The latter result is based on the
Steiner point, which is the exact optimal solution to a related
geometric problem called Lipschitz selection and dates from 1840. In the
talk, I will briefly outline the first solution and fully explain the
second.
Partially based on joint works with S?bastien Bubeck, Bo'az Klartag, Yin
Tat Lee, and Yuanzhi Li.
From marykw at stanford.edu Sun Sep 22 20:17:26 2019
From: marykw at stanford.edu (Mary Wootters)
Date: Sun, 22 Sep 2019 20:17:26 -0700
Subject: [theory-seminar] Robin Pemantle in tomorrow's probability seminar
(9/23)
In-Reply-To:
References:
Message-ID:
Hi all,
See the attached announcement about tomorrow's probability seminar (4pm in
Sequoia 200), which may be interesting to TCS folks.
Speaker: Robin Pemantle, University of Pennsylvania
Title: Recovering a message from a deletion/insertion channel
Best,
Mary
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Sep23-2019.pdf
Type: application/pdf
Size: 88564 bytes
Desc: not available
URL:
From kabirc at stanford.edu Mon Sep 23 08:48:43 2019
From: kabirc at stanford.edu (Kabir Chandrasekher)
Date: Mon, 23 Sep 2019 08:48:43 -0700
Subject: [theory-seminar] Fwd: ISL Colloquium on Thursday Sept 26: Gregory
Valiant, "New Problems and Perspectives on Learning, Sampling, and Memory,
in the Small Data Regime", 4:30-5:30pm, Packard 101
In-Reply-To:
References:
Message-ID:
Hi All,
This Thursday, Professor Valiant will be speaking at the ISL Colloquium.
The talk details are included in the forwarded announcement.
Best,
Kabir
---------- Forwarded message ---------
From: Kabir Chandrasekher
Date: Mon, Sep 23, 2019 at 8:44 AM
Subject: ISL Colloquium on Thursday Sept 26: Gregory Valiant, "New Problems
and Perspectives on Learning, Sampling, and Memory, in the Small Data
Regime", 4:30-5:30pm, Packard 101
To: , <
information_theory_forum at lists.stanford.edu>, <
ee-students-forum at lists.stanford.edu>, <
cs-students-announce at lists.stanford.edu>
Cc: Gregory Valiant
*Title:* New Problems and Perspectives on Learning, Sampling, and Memory,
in the Small Data Regime
*Speaker:* Gregory Valiant (Stanford)
*Time & location:* Thursday September 26 4:30-5:30 pm, Packard 101
Coffee and pastries will be served before the talk at 4pm in the Packard
second floor kitchen
*Abstract:* I will discuss several new problems related to the general
challenge of understanding what conclusions can be made, given a dataset
that is relatively small in comparison to the complexity or dimensionality
of the underlying distribution from which it is drawn. In the first
setting we consider the problem of learning a population of Bernoulli (or
multinomial) parameters. This is motivated by the ``federated learning"
setting where we have data from a large number of heterogeneous
individuals, who each supply a very modest amount of data, and ask the
extent to which the number of data sources can compensate for the lack of
data from each source. Second, I will introduce the problem of data
"amplification". Given n independent draws from a distribution, D, to
what extent is it possible to output a set of m > n datapoints that are
indistinguishable from m i.i.d. draws from D? Curiously, we show that
nontrivial amplification is often possible in the regime where n is too
small to learn D to any nontrivial accuracy. We also discuss connections
between this setting and the challenge of interpreting the behavior of GANs
and other ML/AI systems. Finally (if there is time), I will also discuss
memory/data tradeoffs for regression, with the punchline that any algorithm
that uses a subquadratic amount of memory will require asymptotically more
data than second-order methods to achieve comparable accuracy. This talk is
based on four joint papers with various subsets of Weihao Kong, Brian
Axelrod, Shivam Garg, Vatsal Sharan, Aaron Sidford, Sham Kakade, and Ramya
Vinayak.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From ccanonne at cs.stanford.edu Tue Sep 24 13:51:12 2019
From: ccanonne at cs.stanford.edu (=?UTF-8?Q?Cl=c3=a9ment_Canonne?=)
Date: Tue, 24 Sep 2019 13:51:12 -0700
Subject: [theory-seminar] TCS+ talk: Wednesday, September 25, Mark Sellke,
Stanford
In-Reply-To: <9d41051c-6ae1-4eec-fbd0-b47377552170@cs.stanford.edu>
References: <9d41051c-6ae1-4eec-fbd0-b47377552170@cs.stanford.edu>
Message-ID:
Reminder: this is tomorrow morning! Bagels are not quite convex bodies,
but come learn about the chasing.
Best,
-- Cl?ment
On 9/19/19 8:39 AM, Cl?ment Canonne wrote:
> Hi all,
>
> There will be a TCS+ -- the first of the Fall season -- next Wednesday,
> with Mark Sellke telling us (online, but not from very far) about the
> recent developments in Chasing Convex Bodies. (abstract below)
>
> It'll be Wednesday, September 25th at 10:00 AM in Gates 463A. Breakfast
> at 9:55, talk at 10 on the wall!
>
> Best,
>
> -- Cl?ment
>
>
>
> -------------------------------
> Speaker: Mark Sellke (Stanford)
> Title: Chasing Convex Bodies
>
> Abstract: I will explain our recent understanding of the chasing convex
> bodies problem posed by Friedman and Linial in 1991. In this problem, an
> online player receives a request sequence K_1,...,K_T of convex sets in
> d dimensional space and moves his position online into each requested
> set. The player's movement cost is the length of the resulting path.
> Chasing convex bodies asks if there an online algorithm with cost
> competitive against the offline optimal path. This is both an
> challenging metrical task system and (equivalent to) a competitive
> analysis view on online convex optimization.
>
> This problem was open for d>2 until last year but has recently been
> solved twice. The first solution gives a 2^{O(d)} competitive algorithm
> while the second gives a nearly optimal min(d,sqrt(d*log(T)))
> competitive algorithm for T requests. The latter result is based on the
> Steiner point, which is the exact optimal solution to a related
> geometric problem called Lipschitz selection and dates from 1840. In the
> talk, I will briefly outline the first solution and fully explain the
> second.
>
> Partially based on joint works with S?bastien Bubeck, Bo'az Klartag, Yin
> Tat Lee, and Yuanzhi Li.
>
> _______________________________________________
> theory-seminar mailing list
> theory-seminar at lists.stanford.edu
> https://mailman.stanford.edu/mailman/listinfo/theory-seminar
From bspang at stanford.edu Tue Sep 24 14:32:03 2019
From: bspang at stanford.edu (Bruce Spang)
Date: Tue, 24 Sep 2019 21:32:03 +0000
Subject: [theory-seminar] Two seminars this week!
Message-ID:
Hi all,
Welcome to fall! This week we have TWO theory seminars:
* On Wednesday 9/25 from 3-4pm in Gates 463A, we have David Wajc talking about "Online Matching with General Arrivals?
* On Friday 9/27 from 3-4pm in Gates 463A, we have Omri Ben-Eliezer talking about "Finding monotone patterns in sublinear time?
The abstracts are below. Hope to see you there!
Bruce
Online Matching with General Arrivals
Speaker: David Wajc
The online matching problem was introduced by Karp, Vazirani and Vazirani nearly three decades ago. In that seminal work, they studied this problem in bipartite graphs with vertices arriving only on one side, and presented optimal deterministic and randomized algorithms for this setting. In comparison, more general arrival models, such as edge arrivals and general vertex arrivals, have proven more challenging and positive results are known only for various relaxations of the problem. In particular, even the basic question of whether randomization allows one to beat the trivially-optimal deterministic competitive ratio of 1/2 for either of these models was open. In this paper, we resolve this question for both these natural arrival models, and show the following.
1. For edge arrivals, randomization does not help ? no randomized algorithm is better than 1/2 competitive.
2. For general vertex arrivals, randomization helps ? there exists a randomized (1/2+?(1))-competitive online matching algorithm.
Based on joint work with Buddhima Gamlath, Michael Kapralov, Andreas Maggiori and Ola Svensson at EPFL, to appear in FOCS 2019.
A preprint can be found here:
http://www.cs.cmu.edu/~dwajc/pdfs/gamlath19.pdf
=========
Finding monotone patterns in sublinear time
Speaker: Omri Ben-Eliezer
We consider efficient algorithms for finding monotone subsequences in arrays, that operate in the property testing regime.
Namely, given an array of n real numbers, that is far from free of monotone increasing subsequences of length k, the goal is to find such an increasing subsequence with good probability using as few queries as possible. This is a generalization of monotonicity testing (which corresponds to the case k=2) and is closely related to the classical longest increasing subsequence problem.
In this talk, we shall settle the non-adaptive query complexity of the above problem for any fixed k: the number of non-adaptive queries required to find such an increasing subsequence is $\Theta((\log n)^{\lfloor \log_2 (k) \rfloor})$, where the underlying constant depends only on k and the distance from freeness.
The main ingredient of the proof is an "abundance versus regularity" structural dichotomy that we establish for arrays that are far from free of increasing subsequences. We define a notion of a hierarchical gap profile for subsequences of the array, and show that any such array either contains many long and easy to find increasing subsequences, or it contains a very large collection of disjoint increasing subsequences with the exact same gap profile.
Joint work with C. Canonne, S. Letzter, E. Waingarten, to appear in FOCS'19.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From nehgupta at stanford.edu Wed Sep 25 10:38:42 2019
From: nehgupta at stanford.edu (Neha Gupta)
Date: Wed, 25 Sep 2019 17:38:42 +0000
Subject: [theory-seminar] Theory Lunch 09/26 - Introductions
Message-ID:
Hi everyone,
We will have our first theory lunch of the year on this Thursday (09/26) from noon to 1pm in Gates 463A. We will do introductions this time.
There is still one more talk slot available for the quarter. If you are interested, please sign up in this document -
https://docs.google.com/document/d/1S0QcDMTn-JRaP1cRFRihZyeNfHUmpYLkyORKMcGIBgY/edit
Thanks,
Neha
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From nehgupta at stanford.edu Wed Sep 25 10:38:42 2019
From: nehgupta at stanford.edu (Neha Gupta)
Date: Wed, 25 Sep 2019 17:38:42 +0000
Subject: [theory-seminar] Theory Lunch 09/26 - Introductions
Message-ID:
Hi everyone,
We will have our first theory lunch of the year on this Thursday (09/26) from noon to 1pm in Gates 463A. We will do introductions this time.
There is still one more talk slot available for the quarter. If you are interested, please sign up in this document -
https://docs.google.com/document/d/1S0QcDMTn-JRaP1cRFRihZyeNfHUmpYLkyORKMcGIBgY/edit
Thanks,
Neha
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From bspang at stanford.edu Fri Sep 27 13:56:36 2019
From: bspang at stanford.edu (Bruce Spang)
Date: Fri, 27 Sep 2019 20:56:36 +0000
Subject: [theory-seminar] Two seminars this week!
In-Reply-To:
References:
Message-ID:
Just a reminder that Omri?s talk is at 3pm today in Gates 463A.
Bruce
On Sep 24, 2019, at 2:32 PM, Bruce Spang > wrote:
Hi all,
Welcome to fall! This week we have TWO theory seminars:
* On Wednesday 9/25 from 3-4pm in Gates 463A, we have David Wajc talking about "Online Matching with General Arrivals?
* On Friday 9/27 from 3-4pm in Gates 463A, we have Omri Ben-Eliezer talking about "Finding monotone patterns in sublinear time?
The abstracts are below. Hope to see you there!
Bruce
Online Matching with General Arrivals
Speaker: David Wajc
The online matching problem was introduced by Karp, Vazirani and Vazirani nearly three decades ago. In that seminal work, they studied this problem in bipartite graphs with vertices arriving only on one side, and presented optimal deterministic and randomized algorithms for this setting. In comparison, more general arrival models, such as edge arrivals and general vertex arrivals, have proven more challenging and positive results are known only for various relaxations of the problem. In particular, even the basic question of whether randomization allows one to beat the trivially-optimal deterministic competitive ratio of 1/2 for either of these models was open. In this paper, we resolve this question for both these natural arrival models, and show the following.
1. For edge arrivals, randomization does not help ? no randomized algorithm is better than 1/2 competitive.
2. For general vertex arrivals, randomization helps ? there exists a randomized (1/2+?(1))-competitive online matching algorithm.
Based on joint work with Buddhima Gamlath, Michael Kapralov, Andreas Maggiori and Ola Svensson at EPFL, to appear in FOCS 2019.
A preprint can be found here:
http://www.cs.cmu.edu/~dwajc/pdfs/gamlath19.pdf
=========
Finding monotone patterns in sublinear time
Speaker: Omri Ben-Eliezer
We consider efficient algorithms for finding monotone subsequences in arrays, that operate in the property testing regime.
Namely, given an array of n real numbers, that is far from free of monotone increasing subsequences of length k, the goal is to find such an increasing subsequence with good probability using as few queries as possible. This is a generalization of monotonicity testing (which corresponds to the case k=2) and is closely related to the classical longest increasing subsequence problem.
In this talk, we shall settle the non-adaptive query complexity of the above problem for any fixed k: the number of non-adaptive queries required to find such an increasing subsequence is $\Theta((\log n)^{\lfloor \log_2 (k) \rfloor})$, where the underlying constant depends only on k and the distance from freeness.
The main ingredient of the proof is an "abundance versus regularity" structural dichotomy that we establish for arrays that are far from free of increasing subsequences. We define a notion of a hierarchical gap profile for subsequences of the array, and show that any such array either contains many long and easy to find increasing subsequences, or it contains a very large collection of disjoint increasing subsequences with the exact same gap profile.
Joint work with C. Canonne, S. Letzter, E. Waingarten, to appear in FOCS'19.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From moses at cs.stanford.edu Fri Sep 27 14:24:00 2019
From: moses at cs.stanford.edu (Moses Charikar)
Date: Fri, 27 Sep 2019 14:24:00 -0700
Subject: [theory-seminar] Fwd: [New post] Travel funding for FOCS 2019
In-Reply-To: <28823477.1283.0@wordpress.com>
References: <28823477.1283.0@wordpress.com>
Message-ID:
>From the Theory Matters blog:
FOCS 2019 will be held in Baltimore, MA from
Nov 9-12, 2019. The early registration deadline is October 9th.
For students interested in attending, Shang-Hua Teng has asked me to relay
the message that there is some travel funding available, courtesy of the
NSF. For details, see this website
. Ignore the deadline on that
page, but apply ASAP for full consideration. Women and minorities are
especially encouraged to apply.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From ccanonne at cs.stanford.edu Fri Sep 27 16:21:00 2019
From: ccanonne at cs.stanford.edu (=?UTF-8?Q?Cl=c3=a9ment_Canonne?=)
Date: Fri, 27 Sep 2019 16:21:00 -0700
Subject: [theory-seminar] TCS+ talk: Wednesday, October 2, Shachar Lovett,
UCSD
Message-ID: <47c4886f-8eaf-b590-c277-4a41d969daa1@cs.stanford.edu>
Hi all,
As mentioned at the end of today's seminar, this coming Wednesday,
October 2th, at 10am, Shachar Lovett from UCSD will speak on TCS+ about
"Towards the sunflower conjecture" (abstract below).
I have requested an online seat, meaning that, as usual, the speaker's
slides and face will be moving on the wall of Gates 463A, live, and we
can ask questions (there will also be breakfast).
See you there,
-- Cl?ment
-------------------------------
Speaker: Shachar Lovett (UCSD)
Title: Towards the sunflower conjecture
Abstract: A sunflower with $r$ petals is a collection of $r$ sets so
that the intersection of each pair is equal to the intersection of all.
Erdos and Rado in 1960 proved the sunflower lemma: for any fixed $r$,
any family of sets of size $w$, with at least about $w^w$ sets, must
contain a sunflower. The famous sunflower conjecture is that the bound
on the number of sets can be improved to $c^w$ for some constant $c$.
Despite much research, the best bounds until recently were all of the
order of $w^{cw}$ for some constant c. In this work, we improve the
bounds to about $(log w)^w$.
There are two main ideas that underlie our result. The first is a
structure vs pseudo-randomness paradigm, a commonly used paradigm in
combinatorics. This allows us to either exploit structure in the given
family of sets, or otherwise to assume that it is pseudo-random in a
certain way. The second is a duality between families of sets and DNFs
(Disjunctive Normal Forms). DNFs are widely studied in theoretical
computer science. One of the central results about them is the switching
lemma, which shows that DNFs simplify under random restriction. We show
that when restricted to pseudo-random DNFs, much milder random
restrictions are sufficient to simplify their structure.
Joint work with Ryan Alweiss, Kewen Wu and Jiapeng Zhang.
From moses at cs.stanford.edu Sun Sep 29 22:12:26 2019
From: moses at cs.stanford.edu (Moses Charikar)
Date: Sun, 29 Sep 2019 22:12:26 -0700
Subject: [theory-seminar] Fwd: [isl-colloq] Talk "Generalized Resilience and
Robust Statistics" by Jiantao Jiao (Fr, 11-Oct @ 1:15pm)
In-Reply-To:
References:
Message-ID:
The ISL talk this coming Friday is likely of interest to many folks on this
list.
Cheers,
Moses
---------- Forwarded message ---------
From: ISL Colloquium and Talk Announcements
Date: Sun, Sep 29, 2019 at 7:24 PM
Subject: [isl-colloq] Talk "Generalized Resilience and Robust Statistics"
by Jiantao Jiao (Fr, 11-Oct @ 1:15pm)
To: , isl-colloq <
isl-colloq at lists.stanford.edu>, ee-students-forum <
ee-students-forum at lists.stanford.edu>
Speaker: Jiantao Jiao -- Assistant Professor, UC Berkeley
Title: Generalized Resilience and Robust Statistics
When: Friday, 11-Oct-2019, 1:15pm to 2:15pm
Where: Packard 202
Abstract:
Robust statistics traditionally focuses on outliers, or perturbations
in total variation distance. However, a dataset could be corrupted in
many other ways, such as systematic measurement errors and missing
covariates. We generalize the robust statistics approach to consider
perturbations under any Wasserstein distance, and show that robust
estimation is possible whenever a distribution?s population statistics
are robust under a certain family of friendly perturbations. This
generalizes a property called resilience previously employed in the
special case of mean estimation with outliers. We justify the
generalized resilience property by showing that it holds under moment
or hypercontractive conditions. Even in the total variation case, these
subsume conditions in the literature for mean estimation, regression,
and covariance estimation; the resulting analysis simplifies and
sometimes improves these known results in both population limit and
finite-sample rate. Our robust estimators are based on minimum distance
(MD) functionals (Donoho and Liu, 1988), which project onto a set of
distributions under a discrepancy related to the perturbation. We
present two approaches for designing MD estimators with good finite-
sample rates: weakening the discrepancy and expanding the set of
distributions. We also present connections to Gao et al. (2019)?s
recent analysis of generative adversarial networks for robust
estimation.
Joint work with Banghua Zhu and Jacob Steinhardt
https://arxiv.org/abs/1909.08755
Bio:
Jiantao Jiao is an Assistant Professor in the Department of Electrical
Engineering and Computer Sciences and Department of Statistics at
University of California, Berkeley. He received his B.Eng. degree in
Electronic Engineering from Tsinghua University, Beijing, China in 2012,
and his M.Sc. and Ph.D. degrees in Electrical Engineering from Stanford
University in 2014 and 2018, respectively. He is a recipient of the
Presidential Award of Tsinghua University and the Stanford Graduate
Fellowship. He was a semi-plenary speaker at ISIT 2015 and a co-recipient
of the ISITA 2016 Student Paper Award and MobiHoc 2019 best paper award.
His research interests are in statistical machine learning,
high-dimensional and nonparametric statistics, mathematical programming,
applied probability, information theory, and their applications.
_______________________________________________
isl-colloq mailing list
isl-colloq at lists.stanford.edu
https://mailman.stanford.edu/mailman/listinfo/isl-colloq
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
From bspang at stanford.edu Mon Sep 30 10:19:58 2019
From: bspang at stanford.edu (Bruce Spang)
Date: Mon, 30 Sep 2019 17:19:58 +0000
Subject: [theory-seminar] =?utf-8?q?Dean_Doron_on_=22Nearly_Optimal_Pseudo?=
=?utf-8?q?randomness_From_Hardness=E2=80=9D?=
Message-ID:
Hi all,
For this week's theory seminar, we have Dean Doron talking about "Nearly Optimal Pseudorandomness From Hardness.? It will be on Wednesday 10/2 from 3-4pm in Gates 463A
The abstract is below. Hope to see you there!
Bruce
Nearly Optimal Pseudorandomness From Hardness
Dean Doron
Existing proofs that deduce P=BPP from circuit lower bounds convert randomized algorithms to deterministic ones with a large polynomial slowdown in running time. In this talk, we will see that if we assume exponential lower bounds against nondeterministic circuits, we can convert any randomized algorithm running in time T to a deterministic one running in time T^{2+?} for an arbitrarily small constant ?. Under complexity-theoretic assumptions, such a slowdown is nearly optimal.
Our result follows from a new pseudorandom generator whose construction uses, among other ideas, a new connection between pseudoentropy generators and locally list-recoverable codes.
Joint work with Dana Moshkovitz, Justin Oh and David Zuckerman
-------------- next part --------------
An HTML attachment was scrubbed...
URL: