[Busec] Alison Lewko at MSR on Wed

Leonid Reyzin reyzin at cs.bu.edu
Fri Jan 6 10:58:33 EST 2012


Note Alison Lewko's talk of cryptographic interest next Wed (below).

---------- Forwarded message ----------
From: Irene Money <irenem at microsoft.com>
Date: Fri, Jan 6, 2012 at 10:29 AM
Subject: Microsoft Research New England Weekly Event Digest
To: Leonid Reyzin <reyzin at cs.bu.edu>


Microsoft Research New England Weekly Event Digest

Here is a digest of the upcoming seminars sponsored by the Microsoft
Research New England Lab:


Biology Seminar: From Neural Circuits to Cellular Circuits – Andreas
Pfenning, CMU| Monday, Jan 9 @ 11 AM @ MIT Building 56-114

Biology Seminar: Identifying the Signaling Cascades and Transcriptional
Regulators that Control Stress Responses – Anthony Gitter, CMU | Tues Jan
10 @ 11 AM @ MIT Building 56-614

MSR Seminar: New Proof Techniques for Attribute-Based Encryption – Allison
Bishop Lewko, University of Texas at Austin | Wed Jan 11 @ 11 AM @ MSR

MSR Seminar: Algorithmic solutions to some statistical questions – Gregory
Valiant, Berkeley | Thur Jan 12 @ 11 AM @MSR


ARRIVAL GUIDANCE:
Upon arrival, be prepared to show a picture ID and sign the Building
Visitor Log when approaching the Lobby Floor Security Desk. Alert them to
the name of the MSR event you are attending and ask them to direct you to
the appropriate floor. Typically the talks are located in the First Floor
Conference Center, however sometimes the location may change.

More details follow:

******************************************************************************************************************************
******************************************************************************************************************************
WHO:                          Andreas Pfenning
AFFILIATION:             Duke University Computational Biology and
Bioinformatics
TITLE:                         From Neural Circuits to Cellular Circuits
HOST:                         Ernest Fraenkel
WHEN:                        Monday, January 9
WHERE:                     MIT Building 56-114: See http://whereis.mit.edu/
TIME:                          11 AM – 12 PM
******************************************************************************************************************************

Abstract
Complex behaviors have a large effect on gene expression in the brain
regions involved in the production of that behavior.  In turn, that gene
expression can affect behavior on a larger time scale. To better understand
the relationship between the dynamics of gene expression and complex
behaviors, we take a systems biology approach to the song system in vocal
learning birds.  Song production is a complex (but very quantifiable)
behavior, the neural circuits involved are well studied, and several of the
brain regions involved in that circuit have a strong gene expression
response.  First, we profiled the gene expression program across four song
nuclei in the zebra finch over a time course of seven hours.  We find
hundreds to thousands of genes being regulated as the bird produces learned
vocalizations.  Surprisingly, the majority of gene expression was unique or
strongly enriched in one or a subset of regions.  To understand the
functions and mechanisms in the variation in the response, the gene
expression patterns are related to behavioral variables, such as song
amount and entropy, in a time lagged linear model.  Several behaviors
robustly predict gene expression levels in the particular regions
associated with that behavior.  The tightly controlled regulation of genes
in response to behavior suggests that they may serve a broader role the
nervous system.  We propose a theoretical feed-forward network at the level
of signaling molecules and transcription factors that could function to
make decisions at the cellular level based on a behavioral stimulus.
Connecting the molecular, anatomical, and behavioral levels of a biological
process is difficult challenge in modern neuroscience.  We have developed a
systems biology framework to approach this problem and applied it to the
complex process of producing learned vocalizations.

Biography
Andreas grew up in Pittsburgh and attended Carnegie Mellon University.  He
completed his undergraduate degree in Computer Science there, doing
research on computational models of transcription factor binding sites.
 Now a Ph.D. candidate  in computational biology at Duke, he is the
laboratory of Erich Jarvis in Neurobiology, coadvised by Alexander
Hartemink in Computer Science.  At Duke, Andreas has worked on a variety of
projects including the characterization of the transcription factor CaRF,
the relationship between salt craving and drug addiction, and the evolution
of the vocal learning behavior.

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

WHO:                          Anthony Gitter
AFFILIATION:             Carnegie Mellon
TITLE:                         Identifying the Signaling Cascades and
Transcriptional Regulators that Control Stress Responses
HOST:                         Ernest Fraenkel
WHEN:                        Tuesday, January 10
WHERE:                     MIT Building 56-614: See http://whereis.mit.edu/
TIME:                          11 AM – 12 PM
******************************************************************************************************************************

Abstract
Adaptation to diverse and ever-changing environmental conditions is vital
to the survival of all organisms.  In many stress responses upstream
proteins (e.g. receptors or host proteins that interact with a pathogen)
detect an environmental stimulus and propagate signals via cascades of
protein-protein interactions to the nucleus where transcription factors
selectively activate or inhibit genes.  For many such responses, these
upstream proteins have been at least partially characterized, but knowledge
of the mechanisms through which they transmit information and ultimately
affect transcriptional regulation is much less complete.  We present a
strategy for integrating large-scale data sources to link the signaling and
transcriptional components of environmental stress response.  Our predicted
models include the dynamics of transcription factor activity as well as the
directed pathways that activate these regulators, which are inferred from
the originally undirected protein interaction network.  Studies in model
organisms reveal that our approach can recover known pathways and identify
missing participants, some of which we validated experimentally.
 Furthermore, we discuss clinical applications of our algorithm for
modeling H1N1 influenza infection.

Biography
Anthony Gitter is a Computer Science Ph.D. student in Ziv Bar-Joseph’s
group at Carnegie Mellon University.  His research explores problems in
systems biology using techniques from both machine learning and
combinatorial optimization.  Specific interests include signaling pathways
involved in stress response, biological network redundancy, and dynamic
transcriptional regulatory programs.  Anthony previously interned with
David Heckerman at Microsoft Research Los Angeles, where he worked on
inferring causal associations between genetic variations, gene expression,
and disease.

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

WHO:                          Allison Bishop Lewko
AFFILIATION:             University of Texas at Austin
TITLE:                         New Proof Techniques for Attribute-Based
Encryption
HOST:                         Yael Kalai
WHEN:                        Wed Jan 11
WHERE:                     Microsoft Conference Center located at One
Memorial Drive, First Floor, Cambridge, MA
TIME:                          11 AM – 12 PM
******************************************************************************************************************************

Abstract
In an attribute-based encryption scheme, users are given keys associated to
attributes and ciphertexts are encrypted under access policies describing
which sets of attributes qualify users to decrypt. The goal of such systems
is to provide advanced functionality and flexibility along with strong
security guarantees. We present new techniques for proving full security of
attribute-based encryption schemes. We will also discuss extensions of
these techniques to providing new capabilities, particularly allowing for
multiple authorities to issue keys in a decentralized system. This talk is
based on joint works with Brent Waters and with Tatsuaki Okamoto, Amit
Sahai, Katsuyuki Takashima, and Brent Waters.

Biography
Allison Bishop Lewko received an A.B. degree in mathematics from Princeton
University in 2006 and a Certificate of Advanced Study in Mathematics from
The University of Cambridge in 2007. She is currently a Ph.d. candidate in
the computer science department at The University of Texas at Austin,
advised by Brent Waters. Her current areas of research are cryptography,
distributed computing, harmonic analysis, and combinatorics.

******************************************************************************************************************************
******************************************************************************************************************************
WHO:                          Gregory Valiant
AFFILIATION:             Berkeley
TITLE:                         Algorithmic solutions to some statistical
questions
HOST:                         Adam Kalai
WHEN:                        Thursday, January 12
WHERE:                     Microsoft Conference Center located at One
Memorial Drive, First Floor, Cambridge, MA
TIME:                          11 AM – 12 PM
******************************************************************************************************************************

Abstract
I will discuss three classical statistical problems for which the
computational perspective unlocks insights into the fundamental nature of
these tasks, and suggests new approaches to coping with the increasing size
of real-world datasets.
The first problem is recovering the parameters of a mixture of Gaussian
distributions.  Given data drawn from a single Gaussian distribution, the
sample mean and covariance of the data trivially yield good estimates of
the parameters of the true distribution;  if, however, some of the data
points are drawn according to one Gaussian, and the rest of the data points
are drawn according to a different Gaussian, how can one recover the
parameters of each Gaussian component?  This problem was first proposed by
Pearson in the 1890's, and, in the last decade, was revisited by  computer
scientists.  In a pair of papers with Adam Kalai and Ankur Moitra, we
established that both the sample complexity, and computational complexity
of this problem are polynomial in the relevant parameters (the dimension,
and the inverse of the desired accuracy).
The second problem, investigated in a series of papers with Paul Valiant,
considers the tasks of estimating a broad class of statistical properties,
which includes entropy, L_k distances between pairs of distributions, and
support size.  There are several implications of our results, including
resolving the sample complexity of the `distinct elements problem' (i.e.
given a data table with n rows, how many random rows must one query to
accurately estimate the number of distinct rows?).  We show that on the
order of n/log n rows is both necessary, and sufficient, improving
significantly on both the prior upper and lower bounds for this problem.
Finally I'll describe some new bounds for the problem of learning noisy
juntas (and parities).  Roughly, this problem captures the task of
determining the `relevant' variables---for example, given a large table
with columns representing the expression of many different genes,  and one
final column representing the incidence of some medical condition, how can
one efficiently find the (possibly small) subset of genes that is relevant
to predicting the condition?

Biography
Gregory Valiant is currently finishing his PhD at UC Berkeley under the
supervision of Christos Papadimitriou, and has enjoyed summer internships
at MSR New England, and IBM Almaden.  His research interests include
algorithms, learning theory, evolution, and game theory, with his most
recent work focusing on algorithmic approaches to several classical
questions in statistics.

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

You are currently subscribed to the MSRNE-Seminars mailing list as:
reyzin at cs.bu.edu.

SUBSCRIBING:

Send a blank message to:

join-msrne-seminars at list.research.microsoft.com

UNSUBSCRIBING:

Send a message to your personal unsubscribe address included in each
message, or send a message to:

leave-msrne-seminars at list.research.microsoft.com

or

leave-1300090-242719.46c076656e27395b817aa121af8a4546 at list.research.microsoft.com
Please see our Privacy Statement:
http://research.microsoft.com/en-us/about/listserverprivacy.aspx
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://cs-mailman.bu.edu/pipermail/busec/attachments/20120106/414d18e1/attachment-0001.html 


More information about the Busec mailing list