[PRL] Fwd: Colloquium Digest, Vol 30, Issue 1

[Mitchell Wand]

Some interesting talks coming up at Harvard.  --Mitch

---------- Forwarded message ----------
From: <colloquium-request at seas.harvard.edu>
Date: 2010/4/9
Subject: Colloquium Digest, Vol 30, Issue 1
To: colloquium at seas.harvard.edu


Send Colloquium mailing list submissions to
       colloquium at seas.harvard.edu

To subscribe or unsubscribe via the World Wide Web, visit
       https://lists.deas.harvard.edu/mailman/listinfo/colloquium
or, via email, send a message with subject or body 'help' to
       colloquium-request at seas.harvard.edu

You can reach the person managing the list at
       colloquium-owner at seas.harvard.edu

When replying, please edit your Subject line so it is more specific
than "Re: Contents of Colloquium digest..."

Today's Topics:

  1. Distinguished Lecture in Computational Science    March 24:
     Domain-Specific Languages with Pat Hanrahan (Gioia Sweetland)
  2. CS Colloquium on March 25 - Rob Miller (Gioia Sweetland)
  3. CS Colloquium on April 1 - Fernando Pereira (Gioia Sweetland)
  4. CS Colloquium on April 15 - Nick Feamster (Gioia Sweetland)


---------- Forwarded message ----------
From: Gioia Sweetland <gioia at pacific.harvard.edu>
To: colloquium at seas.harvard.edu
Date: Thu, 18 Mar 2010 13:37:45 -0400
Subject: [Colloquium] Distinguished Lecture in Computational Science March
24: Domain-Specific Languages with Pat Hanrahan
You are cordially invited to the next Distinguished Lecture in
Computational Science, to be given by Pat Hanrahan of Stanford
University.

**********

"Domain-Specific Languages for Heterogeneous Computer Platforms"

Wednesday, March 24, 2010
4:00 p.m.

Room G-115, Maxwell Dworkin, 33 Oxford St., Cambridge

Refreshments served at 3:45 p.m.

*******************************************************
Pat Hanrahan, CANON Professor, Computer Science and Electrical
Engineering Departments, Stanford University

Abstract

Hardware is becoming increasingly specialized because of the need for
power efficiency. One way to gain efficiency is to use high-throughput
processors (e.g. graphics processing units) optimized for data-
parallel applications; these processors deliver more gigaflops per
watt than CPUs optimized for single-threaded programs. Typical
applications, however, consist of both sequential and parallel code
segments. For such applications, the optimal platform will use
heterogenous combinations of different types of processing elements.
Nowadays in high-performance computing, it is common to create hybrid
systems consisting of multi-core CPUs and many-core GPUs combined into
both shared memory multiprocessors and clusters connected by networks.
The challenge is that the computing model has also become more
complicated. A program for a cluster uses MPI, a program for a
symmetric multiprocessing architecture uses threads and locks, and a
program for a GPU uses a data-parallel programming model such as CUDA.
Programs written for one class of machine will not run efficiently on
another class of machines.

Our thesis is that the only practical method for writing programs for
such heterogeneous machines is to raise the level of the programming
model. In particular, we advocate the use of domain-specific languages
(DSLs).  In this talk I will present the case for using DSLs, our work
designing and implementing Liszt (a DSL for solving partial
differential equations on meshes), and our view of the programming
environment needed to create DSLs and to map them to different
platforms. This work is funded by the Stanford DOE PSAAP Center and
the Pervasive Parallelism Laboratory.

About the Speaker

Pat Hanrahan is the CANON Professor of Computer Science and Electrical
Engineering at Stanford University, where he teaches computer
graphics. His current research involves visualization, image
synthesis, virtual worlds, and graphics systems and architectures.
Before joining Stanford, he was a faculty member at Princeton. He has
also worked at Pixar, where he developed volume rendering software and
was the chief architect of the RenderMan Interface--a protocol that
allows modeling programs to describe scenes to high-quality rendering
programs. Professor Hanrahan has received three university teaching
awards. He has received two Academy Awards for Science and Technology,
the Spirit of America Creativity Award, the SIGGRAPH Computer Graphics
Achievement Award, the SIGGRAPH Stephen A. Coons Award and the IEEE
Visualization Career Award. He was recently elected to the National
Academy of Engineering and the American Academy of Arts and Sciences.



---------- Forwarded message ----------
From: Gioia Sweetland <gioia at pacific.harvard.edu>
To: colloquium at seas.harvard.edu
Date: Mon, 22 Mar 2010 14:32:09 -0400
Subject: [Colloquium] CS Colloquium on March 25 - Rob Miller
 Rob Miller of MIT will give a talk entitled *"User Interface Automation
Using Keywords and Pictures"*

Thursday, March 25, 2010
4:00 p.m.
Maxwell Dworkin G-125

Ice cream at 3:30 p.m., Maxwell Dworkin 2nd floor lounge

---------------------------------------------------------------------------------
*
*"User Interface Automation Using Keywords and Pictures"

When users of an application need to customize its behavior or automate a
repetitive task, the usual solution is a scripting interface or API.  Even
assuming a programming interface is provided by the application (not always
the case), learning how to use it still poses problems of complexity and
unfamiliarity to a user accustomed to the application's graphical user
interface.  This talk discusses two novel approaches to application
scripting that control the GUI directly: (1) keyword programming, which uses
keywords to identify GUI controls to operate, resulting in commands like
"click the Search button"; and (2) picture-driven programming, which uses
screenshots of GUI controls, and computer vision techniques to match them
against the GUI.

This talk will discuss some of our explorations into keyword programming and
picture-driven programming on web and desktop applications.  We also
generalize the notion of keyword programming to use keywords for expressing
Java program syntax.  One surprising result is that Java syntax often has
relatively little information content, and can be inferred automatically
from only a handful of keywords -- pointing the way toward programming
systems that reduce the learning and complexity burdens on their users.

Host:  Krzysztof Gajos

Speaker:  Rob Miller is NBX Career Development Associate Professor in MIT's
Department of Electrical Engineering and Computer Science, and a member of
the Computer Science and Artificial Intelligence Laboratory.  He earned his
Ph.D. in Computer Science from Carnegie Mellon University in 2002, and his
dissertation won the CMU SCS Distinguished Dissertation award and earned an
honorable mention in the ACM Distinguished Dissertation competition.  He
received the NSF CAREER award in 2005, and has won four best paper awards at
USENIX and UIST conferences.  His research interests span human-computer
interaction, user interfaces, software engineering, and artificial
intelligence.  His current research focus lies at the intersection of
programming and user interfaces, with the goal of reducing the complexity
barriers that make programming difficult for novices and experts alike.



---------- Forwarded message ----------
From: Gioia Sweetland <gioia at pacific.harvard.edu>
To: colloquium at seas.harvard.edu
Date: Mon, 29 Mar 2010 16:42:06 -0400
Subject: [Colloquium] CS Colloquium on April 1 - Fernando Pereira
 Fernando Pereira of Google will give a talk entitled *"Learning on the Web"
*

Thursday, April 1, 2010
4:00 p.m.
Maxwell Dworkin G-125

Ice cream at 3:30 p.m., Maxwell Dworkin 2nd floor lounge
---------------------------------------------------------------------------------
*
*"Learning on the Web"

It is commonplace to say that the Web has changed everything.  Machine
learning researchers often say that their projects and results respond to
that change with better methods for finding and organizing Web information.
However, not much of the theory or even the current practice, of machine
learning take the Web seriously.  We continue to devote much effort to
refining supervised learning, but the Web reality is that labeled data is
hard to obtain, while unlabeled data is inexhaustible.  We cling to the
assumption that events are drawn independently from a fixed distribution,
while all the Web data generation processes drift rapidly and involve many
hidden correlations.  Many of our theory and algorithms assume data
representations of fixed dimension, while in fact the dimensionality of data
-- for example the number of distinct words in text -- grows as the Web
grows.  While there has been much work recently on learning with sparse
representations, the actual patterns of sparsity in Web data rarely
considered.  Those patterns might be very relevant to the communication
costs of distributed learning algorithms, which are necessary at Web scale,
but little work has been done on this.

Nevertheless, practical machine learning is thriving on the Web.
Statistical machine translation has developed non-parametric algorithms that
learn how to translate by mining the ever-growing volume of source documents
and their translations that are created on the Web.  Unsupervised learning
methods infer useful latent semantic structure from the statistics of term
co-occurrences in Web documents.  Image search achieves improved ranking by
learning from user responses to search results.  In all those cases, Web
scale demanded distributed algorithms.

I will review some of those practical successes to try to convince you that
they are not just engineering feats, but also rich sources of new
fundamental questions that we should be investigating.

Host:  Stuart Shieber

Speaker:  Fernando Pereira is research director at Google.  His previous
appointments include chair of the Computer and Information Science
department of the University of Pennsylvania, head of the Machine Learning
and Information Retrieval department at AT&T Labs, and research and
management positions at SRI International.  He received a Ph.D. in
Artificial Intelligence from the University of Edinburgh in 1982, and he has
120 research publications on natural language processing, machine learning,
speech recognition, bioinformatics, and logic programming as well as several
patents.  He was elected Fellow of the American Association for Artificial
Intelligence in 1991 for his contributions to computational linguistics and
logic programming, and he was president of the Association for Computational
Linguistics in 1993.



---------- Forwarded message ----------
From: Gioia Sweetland <gioia at pacific.harvard.edu>
To: colloquium at seas.harvard.edu
Date: Fri, 09 Apr 2010 15:49:44 -0400
Subject: [Colloquium] CS Colloquium on April 15 - Nick Feamster
 Nick Feamster of Georgia Tech will give a talk entitled *"Network-Level
Spam and Scam Defenses"*

Thursday, April 15, 2010
4:00 p.m.
Maxwell Dworkin G-125

Ice cream at 3:30 p.m., Maxwell Dworkin 2nd floor lounge
---------------------------------------------------------------------------------
*
*"Network-Level Spam and Scam Defenses"

This talk introduces a new class of methods called "behavioral
blacklisting", which identify spammers based on their network-level
behavior.  Rather than attempting to blacklist individual spam messages
based on what the message contains, behavioral blacklisting classifies a
message based on how the message itself was sent (spatial and temporal
traffic patterns of the email traffic itself).  Behavioral blacklisting
tracks the sending behavior of an email sender from a wide variety of
vantage points and establishes "fingerprints" that are indicative of
spamming behavior.  Behavioral blacklisting can apply not only to email
traffic, but also to the network-level behavior of hosting infrastructure
for scam or phishing attacks.  First, I will present a brief overview of our
study of the network-level behavior of spammers. Second, I will describe two
behavioral blacklisting algorithms that are based on insights from our study
of the network-level behavior of spammers.  Finally, I will describe our
ongoing work applying similar behavioral detection techniques to detecting
both online scam hosting infrastructure and phishing attacks.

Host:  Matt Welsh

Speaker:  Nick Feamster is an assistant professor in the College of
Computing at Georgia Tech. He received his Ph.D. in Computer science from
MIT in 2005, and his S.B. and M.Eng. degrees in Electrical Engineering and
Computer Science from MIT in 2000 and 2001, respectively. His research
focuses on many aspects of computer networking and networked systems,
including the design, measurement, and analysis of network routing
protocols, network operations and security, and anonymous communication
systems. He recently received the Presidential Early Career Award for
Scientists and Engineers (PECASE) for his contributions to cybersecurity,
notably spam filtering.
His honors include a Sloan Research Fellowship, the NSF CAREER award, the
IBM Faculty Fellowship, and award papers at SIGCOMM 2006 (network-level
behavior of spammers), the NSDI 2005 conference (fault detection in router
configuration), Usenix Security 2002 (circumventing web censorship using
Infranet), and Usenix Security 2001 (web cookie analysis).


_______________________________________________
Colloquium mailing list
Colloquium at seas.harvard.edu
https://lists.deas.harvard.edu/mailman/listinfo/colloquium
-------------- next part --------------
HTML attachment scrubbed and removed