[Date Prev][Date Next][Date Index]
CSLI Calendar, Wednesday, 20 April 2005, vol. 20:30
CSLI CALENDAR OF PUBLIC EVENTS
______________________________________________________________________
20 April 2005 Stanford Vol. 20, No. 30
______________________________________________________________________
A weekly publication of the
Center for the Study of Language and Information (CSLI)
Stanford University, Cordura Hall, Stanford, CA 94305-4115
http://www-csli.stanford.edu/
____________
ACTIVITIES FROM 20 APRIL 2005 TO 29 APRIL 2005
WEDNESDAY, 20 APRIL 2005
3:45pm Psychology Department Colloquium [20-Apr-05]
Jordan Hall 420:041
"Why the frontal cortex in autism might be talking only to
itself: Local over-connectivity but long-distance disconnection"
Eric Courchesne
UCSD
http://www-psych.stanford.edu/events_colloquium.html
4:00pm UC Berkeley CIS Seminar [20-Apr-05]
Soda Hall 310 (UC Berkeley)
"Mining Associated Text and Images with Dual-Wing Harmoniums"
Eric Xing
Carnegie Mellon University
http://www.eecs.berkeley.edu/~mseeger/cis-seminar
Abstract below
4:00pm UC Berkeley SIMS Distinguished Lecture Series [20-Apr-05]
202 South Hall (Berkeley)
Title to be announced
Xiao Qiang
Director, Berkeley China Internet Project, Journalism, UC Berkeley
http://www.sims.berkeley.edu/events/dls/
4:15pm NLaSP Colloquium [20-Apr-05]
Bldg. 380:380C (Math corner)
"Text analysis as a noisy source/channel decoding problem"
Paul Taylor
Machine Intelligence Lab, University of Cambridge
http://nlp.stanford.edu/events.shtml
Abstract below
4:15pm EE380: Computer Systems Lab Colloquium [20-Apr-05]
Gates B01
"ZettaCore Molecular Memory Technology"
Ritu Shrivastava
ZettaCore, Inc.
http://www.zettacore.com/
http://ee380.stanford.edu/contents.html
7:00pm SCIL Futures of Learning Lecture Series [20-Apr-05]
Wallenberg Hall Learning theater (Bldg. 160)
"Universal Access to all Knowledge"
Brewster Kahle
Internet Archive
http://scil.stanford.edu/
THURSDAY, 21 APRIL 2005
12:15pm Stanford Theory Lunch [21-Apr-05]
Theory Lounge (wing 4B, Gates)
Title to be announced
Sanatan Rai
http://theory.stanford.edu/~mihaela/theorylunch/
4:00pm Personality Lab [21-Apr-05]
Jordan Hall 420:419
"Using Psychology to Reduce Health Disparities Worldwide:
An Internet Health Research Center"
Ricardo Munoz
UCSF, SFGH
http://www-psych.stanford.edu/events_personality.html
4:00pm SRI AI Seminar Series [21-Apr-05]
EJ228, SRI International
"Finding combinatorial-optimization formulations of problems
that are usually attacked with clever heuristics"
Talbot Michael Katz
http://www.ai.sri.com/seminars/
4:00pm PARC Forum [21-Apr-05]
George Pake Auditorium at PARC
"GNU Go"
Daniel Bump
Mathematics, Stanford University
http://www.parc.com/forum/
Abstract below
4:00pm UC Berkeley Cognition, Brain, and Behavior [21-Apr-05]
3105 Tolman (Berkeley)
"Adult Attachment Interview Insecurity, Relatively Impaired
Neuropsychological Performance, and Enhanced Cortisol
Responses to Challenge"
Anne Rifkin
Psychology, UC Berkeley
Ph.D. talk
http://psychology.berkeley.edu/admin/colloquia.html
4:15pm SSP10: Symbolic Systems Forum [21-Apr-05]
Bldg. 380:380C (Math Corner)
"Where the Counterculture Met the New Economy:
Revisiting the WELL and the Origins of Virtual Community"
Fred Turner
Communication, Stanford University
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
Abstract below
4:15pm Information Systems Seminar [21-Apr-05]
Packard 101
"Network Correlated Data Gathering"
Razvan Cristescu
California Institute of Technology
http://isl.stanford.edu/colloquium.html
Abstract below
7:00pm Language and Poetic Form Workshop [21-Apr-05]
Levinthal Hall, Stanford Humanities Center
"Meter: Between Language and Art"
"What's Elision For? Aesthetic Implications of the Meter of
Shakespeare's Sonnets"
Kristin Hanson
English, UC Berkeley
"The Textsetting Problem: An Approach with Stochastic
Optimality Theory"
Bruce Hayes
Linguistics, UCLA
http://www.linguistics.ucla.edu/people/hayes/
http://www.stanford.edu/dept/linguistics/poetics/
7:00pm SDForum Distinguished Speaker Series [21-Apr-05]
George Pake Auditorium at PARC
"The Emerging Economic Paradigm of Open Source"
Bruce Perens
Open Source Cyber Security Policy Research Laboratory,
George Washington University
http://www.sdforum.org/dss/
(Registration required)
FRIDAY, 22 APRIL 2005
12:30pm CS547: Human-Computer Interaction Seminar [22-Apr-05]
Gates B01
"Unified Activity as a Paradigm for Supporting Collaboration"
Tom Moran
IBM Almaden
http://www-pcd.stanford.edu/cs547/
Abstract below
3:00pm Berkeley Information Access Seminar [22-Apr-05]
107 South Hall (Berkeley)
"New Directions for SIMS"
AnnaLee Saxenian.
http://www.sims.berkeley.edu/academics/courses/is296a-1/s05/
3:15pm Friday Cognitive Seminar [22-Apr-05]
Jordan Hall 420:102
title to be announced
Deborah Hendersen
http://www-psych.stanford.edu/events_cognitive.html
3:30pm Linguistics Department Colloquium [22-Apr-05]
Margaret Jacks Hall 460:126
"Stochastic Phonological Knowledge:
General Constraints, Gradient Ranking"
Bruce Hayes
UCLA
http://www.linguistics.ucla.edu/people/hayes/
http://www-linguistics.stanford.edu/colloq/
Abstract below
4:00pm Berkeley Talk [22-Apr-05]
3110 Etcheverry Hall (UC Berkeley)
"Game Theoretical Approaches to Secure and Robust Routing"
Joao Hespanha
Center for Control Engineering and Computation (CCEC), UC Santa Barbara
http://www.ece.ucsb.edu/~hespanha/
Abstract below
MONDAY, 25 APRIL 2005
12:30pm Center for Internet and Society Talk [25-Apr-05]
Law School 180
"Open Architecture as Communications Policy"
Mark Cooper
Consumer Federation of America
http://cyberlaw.stanford.edu/
Abstract below
4:00pm UC Berkeley Linguistics Colloquium [25-Apr-05]
182 Dwinelle Hall (Berkeley)
"Capacities at the Core of Word Learning"
Lori Markson
Psychology, UC Berkeley
http://linguistics.berkeley.edu/events/
Abstract below
4:15pm CS528: Broad Area Colloquium in AI,
Geometry, Graphics, Robotics, and Vision [25-Apr-05]
TCSeq 201
"Animation for Everyone"
Jovan Popovic
MIT
http://graphics.stanford.edu/courses/cs528/
Abstract below
TUESDAY, 26 APRIL 2005
11:00am Intel Berkeley Digital Home Seminar [26-Apr-05]
Intel Research Berkeley, 2150 Shattuck, Penthouse Suite, Berkeley
"Cracking Human Activity Recognition at 10 Chips/Square Meter"
Matthai Philipose
Intel Research
http://www.intel-research.net/berkeley/Seminars.asp
Abstract below
2:45pm SNRC Industry Seminar [26-Apr-05]
Skilling Auditorium
"Challenges at the Intersection of Machine Learning,
Statistical Induction and Systems"
Ira Cohen
Hewlett Packard Labs
http://snrc.stanford.edu/events/industry-seminar/
Abstract below
5:30pm Syntax Workshop [26-Apr-05]
Margaret Jacks 460:126
Title to be announced
Jim McCloskey
Linguistics, UC Santa Cruz
http://ohlone.ucsc.edu/~jim/
http://www-linguistics.stanford.edu/sssg/
9:15pm SSP205: Film Series [26-Apr-05]
Wallenberg 160:314
Pi (1998, 84 mins.)
http://www.stanford.edu/class/symbsys205/
WEDNESDAY, 27 APRIL 2005
12 noon Psychology Developmental Brownbags [27-Apr-05]
Bldg. 380:381U
"Children's reasoning about people as sources of information"
Gail Heyman
UC San Diego
http://www-psych.stanford.edu/events_developmental.html
3:45pm Psychology Department Colloquium [27-Apr-05]
Jordan Hall 420:041
Title to be announced
Liz Phelps
NYU
http://www-psych.stanford.edu/events_colloquium.html
4:00pm Seminar in Algebraic Topology [27-Apr-05]
Bldg. 380:381T (math corner)
"A path to the history of fibrations (with a philosophical aside)"
Jean-Pierre Marquis
University of Montreal, Visiting Fellow SMRC
first of three talks
Abstract below
4:15pm EE380: Computer Systems Lab Colloquium [27-Apr-05]
Gates B01
"Improving the User Experience in Search"
Udi Manber
A9.com
http://ee380.stanford.edu/contents.html
Abstract below
6:00pm Berkeley History and Philosophy of Logic Mathematics, and Science
234 Moses, (Berkeley) [27-Apr-05]
Title to be announced
Alison Gopnik
Psychology, UC-Berkeley
http://math.berkeley.edu/~jhafner/hplm/
THURSDAY, 28 APRIL 2005
12:15pm Stanford Theory Lunch [28-Apr-05]
Theory Lounge (wing 4B, Gates)
Title to be announced
Hovav Shacham
http://theory.stanford.edu/~mihaela/theorylunch/
4:00pm Stanford Algorithms Seminar (AFLB) [28-Apr-05]
Gates 498
Title to be announced
Rohit Khandekar
http://www.cse.iitd.ernet.in/~rohitk/
http://theory.stanford.edu/~aflb/
4:00pm Personality Lab [28-Apr-05]
Jordan Hall 420:419
"Discussion of Measures"
Casey and Bulent
http://www-psych.stanford.edu/events_personality.html
4:00pm UC Berkeley CIS Seminar [28-Apr-05]
Soda Hall 310 (UC Berkeley)
"Parametric Sequence Alignment"
Lior Pachter
Mathematics Department UC Berkeley
http://www.eecs.berkeley.edu/~mseeger/cis-seminar
Abstract below
4:15pm SSP10: Symbolic Systems Forum [28-Apr-05]
Bldg. 380:380C (Math Corner)
"Cognitive Control and Remembering the Past:
Governing Access to Memory"
Anthony Wagner
Psychology, Stanford
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
Abstract below
4:15pm Information Systems Seminar [28-Apr-05]
Packard 101
"Noise distribution free local prediction accuracy of new and
old statistical learning algorithms"
Lee Jones
University of Massachusetts
http://isl.stanford.edu/colloquium.html
Abstract below
FRIDAY, 29 APRIL 2005
12:30pm CS547: Human-Computer Interaction Seminar [29-Apr-05]
Gates B01
"Affective Interaction Paradigms for Animated Characters"
Bill Tomlinson
UC Irvine
http://www-pcd.stanford.edu/cs547/
Abstract below
3:00pm Berkeley Information Access Seminar [29-Apr-05]
107 South Hall (Berkeley)
"Memory Practices in the Sciences"
Geoffrey Bowker
U. of Santa Clara
http://www.sims.berkeley.edu/academics/courses/is296a-1/s05/
Abstract below
3:15pm Friday Cognitive Seminar [29-Apr-05]
Jordan Hall 420:102
"Categories in cognition: Learning to use and using to learn"
Brian Ross
http://www-psych.stanford.edu/events_cognitive.html
3:30pm Stanford Semantics and Pragmatics Workshop [29-Apr-05]
Margaret Jacks Hall 460:126
"Are Intensions Necessary?"
Almerindo Ojeda
UC Davis
http://www-linguistics.stanford.edu/semgroup/
Abstract below
____________
Stanford Blood Center status: Shortage of O+, O-, A+, A-, B+, and B-.
For an appointment: http://bloodcenter.stanford.edu/ or call
650-723-7831. It only takes an hour of your time and you get free
cookies.
____________
UC BERKELEY CIS SEMINAR
on Wednesday, 20 April 2005, 4:00pm-5:00pm
Soda Hall 310 (UC Berkeley)
http://www.eecs.berkeley.edu/~mseeger/cis-seminar
"Mining Associated Text and Images with Dual-Wing Harmoniums"
Eric Xing
Carnegie Mellon University
I will discuss a multi-wing harmonium model for mining multimedia data
that extends and improves on earlier models based on two-layer random
fields, which capture bi-directional dependencies between hidden topic
aspects and observed inputs. This model can be viewed as an undirected
counterpart of the two-layer directed models such as LDA for similar
tasks, but bears significant difference in inference/learning cost
tradeoffs, latent topic representations, and topic mixing mechanisms.
In particular, our model facilitates efficient posterior inference and
robust topic mixing, and potentially provides high flexibilities in
modeling the latent topic spaces. A contrastive divergence and a
variational algorithm are derived for learning. We specialized our
model to a dual-wing harmonium for captioned images, incorporating a
multivariate binomial model for word-counts and a multivariate
Gaussian for color histogram. I will present some empirical results on
the applications of this model to classification, retrieval and image
annotation on news video collections, and show an extensive comparison
with various extant models.
____________
NLASP COLLOQUIUM
on Wednesday, 20 April 2005, 4:15pm - 5:30pm
Bldg. 380:380C (Math Corner)
http://nlp.stanford.edu/events.shtml
"Text analysis as a noisy source/channel decoding problem"
Paul Taylor
Machine Intelligence Lab, University of Cambridge
This talk addresses the issue of what exactly "text" is, and how best
we should model its behaviour. The starting point is a departure from
two widely held views as to what the nature of text or writing
actually is. In the first view, "text" and "words" are confused in
that a sentence of text is seen as comprising a sequence of words.
The view sees text as the written representation of natural language
where writing is thought of as a process of visually recording the
types of things that we say when we speak.
This work proposes a different approach. Firstly, we adopt a noisy
source/channel model which says that words are an input to an encoding
process whose output is noisy text. The primary task then in text
processing is to decode the text and uncover the words in a statistical
manner. The advantage of this technique is a single mechanism which
deals with many of the common text segmentation and disambiguation
tasks, including tokenisation, sentence splitting, POS tagging, spelling
checking and homograph/homonym disambiguation. Secondly, we challenge
the assumption that text encodes natural language. Rather we see text as
a written encoding of a number of different semiotic systems of which
language is just one; hence we we encounter numbers, dates, mathematics
and email addresses, we should adopt a similar but parallel approach to
that used for "real" natural language.
____________
PARC FORUM
on Thursday, 21 April 2005, 4:00pm - 5:00pm
George Pake Auditorium, PARC
http://www.parc.com/forum/
"GNU Go"
Daniel Bump
Mathematics, Stanford University
GNU Go is a computer program that plays Go. It is the only reasonably
strong program that is non-proprietary. A look at GNU Go internals
will be given, showing how GNU Go analyzes situations from actual
play. GNU Go is a project of the Free Software Foundation. For more
information about GNU Go see
http://www.gnu.org/software/gnugo/devel.html
About the Speaker: Daniel Bump's research is in automorphic forms,
representation theory, and number theory.
____________
SSP10: SYMBOLIC SYSTEMS FORUM
on Thursday, 21 April 2005, 4:15pm
Bldg. 380:380C (Math Corner)
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
"Where the Counterculture Met the New Economy:
Revisiting the WELL and the Origins of Virtual Community"
Fred Turner
Communication, Stanford University
Over the last ten years, scholars have largely ascribed the rise of
virtual community to the widespread adoption of computer networking
technologies. This paper examines the history of the system on which
the term "virtual community" was first used, the Whole Earth Lectronic
Link (or WELL), and shows that as both an idea and a social formation,
virtual community in fact emerged at the intersection of three forces:
the appearance of public computer networks, the persistence of
countercultural social ideals from the 1960s, and a shift toward
networked forms of economic activity. In the process, the paper brings
together analytical frameworks from organizational sociology, American
cultural history, and science and technology studies in order to
illuminate the complex ways in which technological, social and
cultural forms co-evolve.
About the Speaker: Fred Turner is an Assistant Professor of
Communication at Stanford University. He is the author of
Counterculture Into Cyberculture: How Stewart Brand and the Whole
Earth Network Transformed the Politics of Information (forthcoming,
University of Chicago Press) and Echoes of Combat: The Vietnam War in
American Memory (Anchor Books, 1996; 2nd ed. University of Minnesota
Press, 2001).
____________
INFORMATION SYSTEMS SEMINAR
on Thursday, 21 April 2005, 4:15pm-5:15pm
Packard 101
http://isl.stanford.edu/colloquium.html
"Network Correlated Data Gathering"
Razvan Cristescu
California Institute of Technology
We consider the problem of correlated data gathering by a network with
a sink node and a tree communication structure, where the goal is to
minimize the total transmission cost of transporting the information
collected by the nodes, to the sink node. We analyze two lossless
coding strategies: a Slepian-Wolf model where optimal coding is
complex and transmission optimization is simple, and a joint entropy
coding model with explicit communication where coding is simple and
transmission optimization is difficult. For the Slepian-Wolf setting,
we derive a closed form solution and propose an efficient distributed
approximation algorithm; we generalize our results to the case of
multiple sinks. For the explicit communication case, we prove, by a
non-trivial reduction from the min-set cover problem, that building an
optimal data gathering tree is NP-hard; we propose various efficient
distributed approximation algorithms. Further, we compare
asymptotically, for dense networks, the total costs associated with
Slepian-Wolf coding and explicit communication by finding their
corresponding scaling laws and analyzing the ratio of their respective
costs. We show that under certain conditions on the correlation
structure, Slepian-Wolf coding provides unbounded gains in terms of
transmission efficiency over the straightforward approach of
opportunistic aggregation and compression by explicit communication.
Next, we address the related problem of finding optimal efficient
sensor placement to improve both the total energy consumption and the
network lifetime in a sensor network.
Also, we study data gathering for the case of high-resolution lossy
coding, under distortion constraints. We extend our framework to the
case of deterministic signals by designing efficient wavelet-based
decentralized multiresolution algorithms that minimize communication
costs. Finally, we show that for spatio-temporal random processes
there exists an optimal finite network density that minimizes the
total distortion of data gathering.
____________
CS547: HUMAN-COMPUTER INTERACTION SEMINAR
on Friday, 22 April 2005, 12:30-2:00pm
Gates B01
http://www-pcd.stanford.edu/cs547/
"Unified Activity as a Paradigm for Supporting Collaboration"
Tom Moran
IBM Almaden Research Center
http://hci-journal.com/moran/
We use the term "Business Activity" to encompass not only
formally-specified in business processes, but also the situational,
semi-organized, on-the-ground work that people do together. The former
are "run" by workflow-driven business applications and the latter are
supported by standard collaboration tools. Systems deal with tools and
artifacts; activities are mostly tacit, being only partially
represented by ad hoc, localized, and specialized constructs. We argue
that the full range of business activity should be supported in
business platforms by an explicit construct we call "Unified
Activity," which will enable a new degree of business integration,
easier collaboration, capture of business practice, and more effective
business process evolution. The goal of Unified Activity is to create
a unified representation for Business Activities, both informal and
formal, that supports the aggregation of resources to support the
activities, interaction with formal business process workflows, and
the creation and evolution of activity patterns. This talk will
discuss our efforts to define Unified Activity as an integration focus
for IBM software platforms, principally Lotus Workplace. We will
describe the representation and supporting architecture, show the user
experience in various client applications, and discuss the research
challenges for impacting business practices.
About the Speaker: Tom Moran pioneered the establishment of the field
of human-computer interaction (HCI) within computer science. He was
Principal Scientist and manager of the User Interface and the
Collaborative Systems Areas at Xerox PARC (1974-2001) and was founding
Director of Xerox EuroPARC in Cambridge, England (1986-90).
He is now a Distinguished Engineer at IBM. He founded the influential
journal Human Computer Interaction in 1984 and continues as its
Editor. He is an ACM Fellow and recipient of ACM SIGCHI's Lifetime
Achievement Award. After graduating from CMU, Tom worked with Allen
Newell and Stu Card at Xerox PARC on the theoretical foundations of
human-computer interaction, which culminated in their seminal book,
The Psychology of Human-Computer Interaction (1983). Their Model Human
Processor, Keystroke-Level Model, and GOMS Model have influenced a
generation of HCI researchers. Tom worked with Xerox designers in the
1970s to formulate the design methodology for the Xerox Star user
interface, the first "desktop metaphor." His analytic research, in
addition to the psychology of HCI, includes the Command Language
Grammar, task mapping and user conceptual models, the workaday world
paradigm for CSCW, design rationale, and embodied user interfaces. His
systems design work includes the NoteCards idea-processing hypertext
system, the user-tailorable Buttons system, the RAVE media space, the
Tivoli electronic whiteboard, multimedia meeting capture and
"salvaging" tools, whiteboard-embedded meeting tools, and
camera-captured walls. At IBM, he is leading a multi-lab research
program on Unified Activity Management with the Lotus division.
____________
LINGUISTICS DEPARTMENT COLLOQUIUM
on Friday, 22 April 2005, 3:30pm
Margaret Jacks Hall 460:126
http://www-linguistics.stanford.edu/colloq/
"Stochastic Phonological Knowledge:
General Constraints, Gradient Ranking"
Bruce Hayes
UCLA
http://www.linguistics.ucla.edu/people/hayes/
Recent research has documented an ability of language learners to
project statistical knowledge from the lexicon. Where the language
contains competing morphological patterns (for example iU ~ +U vs. X ~
Xd past tenses in English, or competing verbal conjugation classes in
Romance) language learners become tacitly aware of the relative
lexical frequencies of the rival patterns. Supporting evidence comes
from experiment: asked to inflect novel stems compatible with more
than one pattern, speakers behave stochastically, adopting one or
another pattern at random in frequencies matching the lexical
frequencies.
This ability has been shown for opaque alternations (Zuraw 2001),
conjugation class choice (Albright 2002), and projection of underlying
representations (Ernestus and Baayen 2003). The present work extends
this research line to a classical case of natural, transparent
phonology, the vowel harmony system of Hungarian. A formal analysis
using stochastic Optimality Theory is proposed, and it is demonstrated
that the rankings are learnable using a combination of algorithms from
the literature.
I conclude with data suggesting, tentatively but intriguingly, that
language learners favor general that is, formally simple constraints.
When the Hungarian system is modeled using constraints that are too
detailed, the result is a better match to the training set, but a
worse match to native speaker intuitions. The suggested conclusion is
that the speakers learn a grammar that diverges somewhat from the
input data pattern, in the direction of greater generality.
____________
BERKELEY TALK
on Friday, 22 April 2005, 4:00pm - 5:00pm
3110 Etcheverry Hall (UC Berkeley)
"Game Theoretical Approaches to Secure and Robust Routing"
Joao Hespanha
Center for Control Engineering and Computation (CCEC), UC Santa Barbara
http://www.ece.ucsb.edu/~hespanha/
At any given point in time, traditional Internet routing uses a single
path to forwards data packets from a source to a destination. However,
there are advantages to deviating from this principle. In multi-path
routing, packets are spread over several distinct paths in their way
from a source to a destination. The use of multiple paths has been
proposed to increase network security and improve fault tolerance, as
well as to increase data throughput, reduce traffic congestion, and
improve network utilization.
This talk addresses the computation of multi-path routing tables so as
to realize the full potential of this type of network routing. We
frame routing as a mathematical game between a player that forwards
packets along the network and an adversary that attempts to prevent
data from reaching its destination. The equilibria (saddle) solutions
to these games provide stochastic multi-path routing tables that have
several desirable properties in terms of security, robustness, and
load balancing.
Two crucial obstacles to the broad use of multi-path path routing are
the computational complexity of computing optimal routing tables and
its impact on TCP congestion control. In this talk we report on recent
progress in addressing these issues: Computational complexity can be
mitigated though hierarchical decompositions of the routing problem
and new versions of TCP have been developed to overcome the severe
performance degradation that this protocol exhibits under persistent
packet re-ordering, which is typically observed under multi-path
routing.
About the Speaker: Joao P. Hespanha was born in Coimbra, Portugal. He
received the Licenciatura and the M.S. degree in electrical and
computer engineering from Instituto Superior Tecnico, Lisbon,
Portugal, in 1991 and 1993, respectively, and the M.S. and
Ph.D. degrees in electrical engineering and applied science from Yale
University, New Haven, Connecticut, in 1994 and 1998,
respectively. For his PhD work, Dr. Hespanha received Yale
University's Henry Prentiss Becton Graduate Prize for exceptional
achievement in research in Engineering and Applied
Science. Dr. Hespanha is the author of over 100 technical papers, the
recipient of a NSF CAREER Award, and the PI in several federally
funded projects. His research interests include hybrid and switched
systems; the modeling and control of communication networks;
distributed control over communication networks (also known as
networked control systems); the use of vision in feedback control; and
the control of haptic devices. He is currently an Associate Editor for
the IEEE Transactions on Automatic Control.
____________
CENTER FOR INTERNET AND SOCIETY TALK
on Monday, 25 April 2005, 12:30pm - 1:30pm
Law School 180
http://cyberlaw.stanford.edu/
"Open Architecture as Communications Policy"
Mark Cooper
Director of Research, Consumer Federation of America
The ability of service providers, application developers and hardware
firms to innovate and consumers to communicate over the Internet will
be determined in two high profile Supreme Court cases in the current
term - National Cable Telecommunications Association and Federal
Communications Commission v. Brand X and Metro-Goldwyn-Mayer Studies,
Inc. V. Grokster. In Open Architecture as Communications Policy, Dr.
Cooper shows that policymakers have failed to appreciate how
profoundly important open networks and the flow of information have
been for dynamic innovation in the digital economy.
About the Speaker: Dr. Cooper holds a Ph.D. from Yale University and
is a former Yale University and Fulbright Fellow. He is Director of
Research at the Consumer Federation of America, a Fellow at the
Stanford Law School Center for Internet and Society, the McGanon
Communications Research Center, and an Associated Fellow at the
Columbia University Institute on Tele-Information.
He combines an active practice as an expert witness and an advocate
with academic activities. He has provided expert testimony in over 250
cases for public interest clients including Attorneys General,
People's Counsels, and citizen interveners before city, state and
federal agencies, courts and legislators in almost four dozen
jurisdictions in the U.S. and Canada. He is the author of Books - Open
Architecture as Communications Policy (Center for Internet and
Society, 2004), Media Ownership and Democracy in the Digital
Information Age: Promoting Diversity with First Amendment Principles
and Market Structure Analysis (Center for Internet and Society, 2003),
Cable Mergers and Monopolies: Market Power in Digital Media and
Communications Networks (Economic Policy Institute, 2002), Equity and
Energy (Westview, 1983), The Transformation of Egypt (Johns Hopkins,
1982).
Chapters in edited volumes - "Building A Progressive Media And
Communications Sector," forthcoming In News Incorporated: Corporate
Media Ownership And Its Threat To Democracy, Elliot Cohen, Ed.
Prometheus Books; "Hyper-Commercialism In The Media: The Threat To
Journalism And Democratic Discourse," forthcoming In Converging Media,
Diverging Politics: A Political Economy Of News In The United States
And Canada, Snyder-Gasher-Compton-Eds. Lexington Books; "Reclaiming
The First Amendment: Legal, Factual And Analytic Support For Limits On
Media Ownership," forthcoming in The Future of Media, Robert
McChesney, Ed., Seven Stories Press; "The Digital Divide Confronts the
Telecommunications Act of 1996: Economic Reality versus Public
Policy," in B.M. Compaine (Ed.) The Digital Divide Cambridge: MIT,
2001)
Articles in trade and scholarly journals including law review articles
on media, telecommunications and digital society issues - "Open
Communications Platforms: Cornerstone Of Innovation And Democratic
Discourse In The Internet Age, The Journal of Telecommunications and
High Technology Law, 2003; "Inequality in Digital Society," Cardozo
Arts and Entertainment Law Journal, 2002; "Antitrust as Consumer
Protection in the New Economy: Lessons From the Microsoft Case,"
Hasting Law Journal, April 2001; and "Open Access to the Broadband
Internet," University of Colorado Law Review, Fall 2000.
____________
UC BERKELEY LINGUISTICS COLLOQUIUM
on Monday, 15 April 2005, 4:00pm
182 Dwinelle Hall (Berkeley)
http://linguistics.berkeley.edu/events/
"Capacities at the Core of Word Learning"
Lori Markson
Psychology, UC Berkeley
Children's striking ability to learn new words quickly and efficiently
has spurred scholars to pinpoint the mechanism its core. An ongoing
controversy has concerned whether word learning results from a
domain-specific or domain-general mechanism. A current prevalent view
is that children utilize multiple cues when acquiring new word
meanings, suggesting a host of capacities may be influential in this
process. Such capacities include general cognitive abilities of
attention and memory, lexical-specific constraints, and sensitivity to
the intentions of others. An important question that arises is the
relative influence of these different cues on the word learning
process. My colleagues and I have addressed this question by comparing
children's learning of words and non-linguistic information, and
examining the impact of children's understanding of communicative
intent. Taken together, the findings from diverse laboratories support
the notion that word learning is the result of a mechanism that is not
specific to language. Rather, some cognitive capacities that are in
place and used for other purposes, are recruited by humans for the
task of learning words.
____________
CS528: BROAD AREA COLLOQUIUM FOR
AI-GEOMETRY-GRAPHICS-VISION-ROBOTICS
on Monday, 25 April 2005, 4:15pm
TCSeq 201
http://graphics.stanford.edu/courses/cs528/
"Animation for Everyone"
Jovan Popovic
MIT
Computer animation can revolutionize standard practices in education
and training by enabling inexpensive, high-quality computer-based
instruction. Unfortunately, standard animation techniques require a
tremendous amount of artistry, skill, and time: animations with
realistic motions require manual selection of appropriate simulation
parameters; and animations with highly stylized actions require manual
deformations of geometric shapes. This talk describes our recent
progress towards simplifying animation tasks through acquisition,
analysis, and modeling of human motion and shape deformation. We plan
to incorporate these solutions into rapid-animation tools with the
hope of making animation as easy as speaking into a microphone or
typing in a word processor.
About the Speaker: Jovan Popovic is an assistant professor in the
Department of Electrical Engineering and Computer Science and a member
of the Computer Graphics Group in the Computer Science and Artificial
Intelligence Laboratory. Before joining MIT in the Fall of 2001, Jovan
Popovic received his Ph.D. in Computer Science from Carnegie Mellon
University and his B.S. degrees in Mathematics and Computer Science
from Oregon State University.
____________
INTEL BERKELEY DIGITAL HOME SEMINAR
on Tuesday, 26 April 2005, 11:00am - 12:30pm
Intel Research Berkeley, 2150 Shattuck, Penthouse Suite, Berkeley
http://www.intel-research.net/berkeley/Seminars.asp
"Cracking Human Activity Recognition at 10 Chips/Square Meter"
Matthai Philipose
Intel Research
A system that reasons about what day-to-day activity a person is
performing has been a long-standing but elusive goal of automated
perception. Given the large variety of daily activities and the
environments in which they are performed, it is unclear what features
to sense, how to sense them, how to relate these features in a model,
how to obtain the models for each activity and how to reason
appropriately using these models. Pending solutions to these issues,
state of the art systems have been restricted to recognizing small
numbers of activities performed in relatively constrained ways. In
this talk, I will present the System for Human Activity Recognition
and Prediction (SHARP) developed at Intel Research Seattle. SHARP
promises to be able to recognize large numbers (likely thousands) of
activities in practice. I will focus on four key technologies that
distinguish SHARP. First, I will show how a novel class of sensors we
have developed (based on Radio Frequency Identification) can be
deployed at very high density and used to provide extremely detailed
information on what objects are in use at any given time. Second, I
will show that given these features, many activities can be described
effectively using particularly simple probabilistic models. Third, I
will show how the simplicity of these models, together with the
genericity of daily activities, allows the models to be mined from
online common sense sources such as the web with absolutely no human
supervision per activity; the generic models can then be customized to
each person using simple machine learning techniques. Fourth, I will
show how to use such models to rate how well activities are performed,
and to provide suggestions on how to improve the performance when
appropriate. I will analyze the effectiveness of these techniques
using a variety of metrics. As an overall measure, in an experiment
with 14 subjects performing roughly 50 activities ranging from
brushing teeth to making a sandwich in a real home, SHARP was able to
detect activities that occurred 73% of the time, with few false
positives. This is joint work with Tanzeem Choudhury, Sunny Consolvo,
Ken Fishkin, Dirk Haehnel and Josh Smith of Intel Research Seattle,
and collaborators from the University of Washington, Carnegie Mellon
University and amazon.com.
____________
SNRC INDUSTRY SEMINAR
on Tuesday, 26 April 2005, 2:45pm - 4:00pm
Skilling Auditorium
http://snrc.stanford.edu/events/industry-seminar/
"Challenges at the Intersection of Machine Learning,
Statistical Induction and Systems"
Ira Cohen
Hewlett Packard Labs
Recent research activity has shown encouraging results for performance
debugging and failure diagnosis and detection in systems by using an
approach based on automatically inducing models and correlating
data. In this talk I will explore research questions and preliminary
results regarding the next steps to move this line of research
further. I will specifically formulate three challenges. The first one
concerns the need to ensure the validity of these models at all
times. This challenge encompasses research issues ranging from model
evaluation and diagnosis to the management of the models lifecycle
with the aim of achieving online adaptation to system changes. The
second challenge deals with the interaction with human operators. It
includes issues about generating explanations, enabling feedback,
handling of false positives and missed detections. The third challenge
focuses on how to leverage previously diagnosed/repaired problems and
the historical behavior of the system. Specifically I will discuss
methods for transforming the output of probabilistic models into
machine structured representations or signatures of the system state,
so that they serve as index for diagnosis and repairs.
About the Speaker: Ira Cohen is a researcher at Hewlett Packard
research labs, where he works on applying machine learning and pattern
recognition techniques to system diagnosis, management and
control. Ira joined HP-Labs in 2003 after receiving his PhD in
Electrical and Computer Engineering from the University of Illinois at
Urbana-Champaign. His research interests are in probabilistic models,
systems management and control, computer vision and human computer
interaction.
____________
SEMINAR IN ALGEBRAIC TOPOLOGY
on Wednesday, 27 April, 4 May, 11 May 2005, 4:00pm
Bldg. 380:381T (math corner)
"A path to the history of fibrations (with a philosophical aside)"
Jean-Pierre Marquis
University of Montreal, Visiting Fellow SMRC
In this series of talks, I will look at the history of the concept of
fibration in homotopy theory. My goal is to show that the history of
the concept of fibration illustrates important aspects of the
development and the nature of mathematical knowledge, particularly the
development and the nature of mathematics of the 20th century. I will
first sketch the elements of the history of homotopy theory itself,
then I will concentrate on the history of the concept of fibration,
from Hurewicz & Steenrod's definition of fiber space to Quillen's
model categories. I will argue that the concept of fibration is valued
by mathematicians for reasons that are not usually recognized by
philosophers of mathematics and that, moreover, these values force us
to recognize that there is, within pure mathematics, a type of
knowledge which is akin to a form of systematic technology. If this is
correct, I believe we should rethink the whole organization of
mathematical knowledge.
____________
EE380: COMPUTER SYSTEMS LAB COLLOQUIUM
on Wednesday, 27 April 2005, 4:15pm to 5:30pm
B01, Gates Computer Science Building
http://ee380.stanford.edu/contents.html
"Improving the User Experience in Search"
Udi Manber
A9.com
http://a9.com/
I'll highlight several new projects that A9.com has launched recently,
including BlockView (our Yellow Pages with images) and OpenSearch (an
extension to RSS to allow people to syndicate search), and talk in
general about the issues of improving the user experience in search
beyond the usual "enter 2-3 words and get a list of links".
About the speaker: Udi Manber was a professor of Computer Science at
the University of Arizona where he co-developed several popular search
software packages, including Agrep, Glimpse, and Harvest. He left
academia to join Yahoo as Chief Scientist in 1998. In 2002 he moved to
Amazon.com with the unusual title of Chief Algorithms Officer, where
he worked on Amazons Search Inside the Book project, among many other
things, and in Nov 2003 he became the CEO of A9.com, a subsidiary of
Amazon.com developing innovative search technologies.
____________
UC BERKELEY CIS SEMINAR
on Thursday, 28 April 2005, 4:00pm-5:00pm
Soda Hall 310 (UC Berkeley)
http://www.eecs.berkeley.edu/~mseeger/cis-seminar
"Mining Associated Text and Images with Dual-Wing Harmoniums"
Lior Pachter
Mathematics, UC Berkeley
One of the major successes in computational biology has been the
unification, using the graphical model formalism, of a multitude of
algorithms for annotating and comparing biological sequences.
Graphical models that have been applied towards these problems include
hidden Markov models for annotation, tree models for phylogenetics,
and pair hidden Markov models for alignment. A single algorithm, the
sum-product algorithm, solves many of the inference problems
associated with different statistical models. We will discuss the
polytope propagation algorithm for computing the Newton polytope of an
observation from a graphical model. This algorithm is a geometric
version of the sum-product algorithm and is used to analyze the
parametric behavior of maximum a posteriori inference calculations for
graphical models. We will focus on the applications to sequence
alignment.
These results will appear in a contributed chapters (by Colin Dewey,
Sergi Elizalde, Michael Joswig, Radu Mihaescu and Kevin Woods) of a
forthcoming book "Algebraic Statistics for Computational Biology"
(joint with Bernd Sturmfels).
____________
SSP10: SYMBOLIC SYSTEMS FORUM
on Thursday, 28 April 2005, 4:15pm
Bldg. 380:380C (Math Corner)
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
"Cognitive Control and Remembering the Past: Governing Access to Memory"
Anthony Wagner
Psychology Department, Stanford University
Remembering the past is not a unitary cognitive act, but rather
depends on an ensemble of processes, including cognitive control
mechanisms that guide access to goal-relevant memories. In this talk,
evidence will be provided for a distinction between two basic forms of
cognitive control--top-down (controlled) retrieval of potentially
relevant representations and post-retrieval selection. Initial data
will highlight the role of selection and controlled retrieval
mechanisms during (a) interference resolution in working memory, (b)
retrieval of task-relevant semantic knowledge, and (c) switching
between task sets. Subsequently, the role of retrieval and selection
for remembering episodic details about the past will be discussed. As
such, this talk aims to illustrate how core mechanisms of cognitive
control support flexible behavior across a range of memory domains.
____________
INFORMATION SYSTEMS SEMINAR
on Thursday, 28 April 2005, 4:15pm-5:15pm
Packard 101
http://isl.stanford.edu/colloquium.html
"Noise distribution free local prediction accuracy
of new and old statistical learning algorithms"
L. K. Jones
University of Massachusetts Lowell
Local learning is the process of determining the value of an unknown
function at only one fixed query point based on information about the
values of the function at other points. We propose an optimal
methodology for local learning which differs from ( and is
demonstrated in many interesting cases to be superior to) several
popular local and global learning methods. In this theory the
objective is to minimize the (maximum) prediction error at the query
point only - rather than minimize some average performance over the
entire domain of the function. Since different compute-intensive
procedures are required for each different query, local learning
algorithms have only recently become feasible due to the advances in
computer availability, capability and parallelizability of the last
two decades.
In this talk we will first explain with highway profiling data in two
dimensions how the theory works for simple regression and compare it
to smoothing splines. Extensions to inverse problems in one dimension
are discussed and a computationally challenging method is proposed for
tomography. New and old algorithms with optimal local prediction error
bounds for ridge and lasso regression will then be presented. Using
these bounds an optimal local aggregate estimator is derived from the
trees in a random forest. Finding the estimator requires the solution
to a challenging large dimensional non-differentiable convex
optimization problem. Approximate solutions to the forest
optimizations will be given for classification using micro-array data.
____________
CS547: HUMAN-COMPUTER INTERACTION SEMINAR
on Friday, 29 April 2005, 12:30-2:00pm
Gates B01
http://www-pcd.stanford.edu/cs547/
"Affective Interaction Paradigms for Animated Characters"
Bill Tomlinson
UC Irvine
http://www.ics.uci.edu/~wmt/
This talk describes research into novel interaction paradigms that
help people engage with interactive animated characters. Engaging
character-based interactions can provide new platforms for education
and entertainment. Several past and current projects will be shown,
including Alan Alda howling at the AlphaWolf installation, and the
Virtual Raft Project from the Interactivity program at CHI 05.
About the Speaker: Bill Tomlinson is an Assistant Professor of
Informatics and Drama at the University of California, Irvine, where
he teaches in the ACE (Arts Computation Engineering) graduate
program. He is a researcher and animator of autonomous computational
characters, and a designer of interaction paradigms that enable people
to engage with these characters. Previous interactive projects have
been shown at SIGGRAPH, Ars Electronica, the Game Developers
Conference, the ZKM Future Cinema exhibition and other venues, and
have been reviewed by CNN, the Wall Street Journal, Sculpture
Magazine, Scientific American Frontiers, the LA Times, Wired.com and
the BBC. His 1996 animated film, Shaft of Light, screened at the
Sundance Film Festival and was distributed by the Anti-Defamation
League in its Anti-Bias/Diversity Catalog. He holds an A.B. in Biology
from Harvard College, an M.F.A. in Experimental Animation from
CalArts, and S.M. and Ph.D. degrees from the MIT Media Lab.
____________
BERKELEY INFORMATION ACCESS SEMINAR
on Friday, 29 April 2005, 3:00pm - 5:00pm
107 South Hall (Berkeley)
http://www.sims.berkeley.edu/academics/courses/is296a-1/s05/
"Memory Practices in the Sciences"
Geoffrey Bowker
U. of Santa Clara
Lev Manovich has argued that the database is the central symbolic
form of our times. In this paper, I take a long history of databases and
databasing technologies over the past two hundred years, discussing the
organizational and social dimensions of working with a new information
infrastructure. At the same time, I explore the consequences of this
change for the kinds of stories that scientists get to tell about the
past - be this of human history, the history of our species or the
history of our planet. Finally, I discuss how the proclaimed development
of a new global cyberinfrastructure for science both inflects and is
affected by current work practice.
____________
STANFORD SEMANTICS AND PRAGMATICS WORKSHOP
on Friday, 29 April 2005, 3:30pm
Margaret Jacks Hall 460:126
http://www-linguistics.stanford.edu/semgroup/
"Are Intensions Necessary?"
Almerindo Ojeda
University of California-Davis
Possible-worlds semantics (PWS) reaches a well-known impasse as it
attempts to interpret expressions that have the same intension yet
differ in meaning. To overcome this impasse, PWS is routinely
supplemented with structural theories of meaning. According to these
theories, the meaning of an expression e is not just its intension e',
but rather a structure of intensions--say, a tree--that reflects the
way in which e' was built, compositionally, from simpler intensions.
It is generally agreed that structural theories of meaning are strong
enough to carry PWS over the wall of nonsynonymous cointensionals. It
thus comes as a surprise that considerations of simplicity did not
lead us to ask whether structural theories of meaning were not strong
enough to carry all of the weight PWS lifted--or whether structural
theories of meaning could not supplant rather than supplement PWS.
This is the question I will address in this talk. To do so I develop a
purely extensional structural theory (PEST) of meaning. According to
this PEST, the meaning of an expression e is not just its extension
e', but rather a structure of extensions (say, a tree) that reflects
the way in which e' was built, compositionally, from simpler
extensions. I then show that, coupled with independently-motivated
proof-theoretic interpretations of modality and counterfactuals, PESTs
can afford satisfying solutions to the problems that drove extensional
semantics into the ground--and can do so, of course, without appealing
to the vast intractable ontology of possible worlds. The talk closes
with replies to plausible objections that might be levelled against
PESTs. For, morphosyntactic structure underdetermines meaning. And
overdetermines it as well. Even more fundamentally, like all
structural theories of meaning, PESTs clash with the set-theoretic
Axiom of Well-Foundedness.
____________
END MATERIAL
The CSLI Calendar appears weekly on most Wednesdays throughout the
year. Announcements, abstracts, and other information to appear in
the Calendar should be submitted to the editor, who reserves the right
to decide what does or does not go in the calendar
mailto:incalendar@csli.stanford.edu
Requests to be added to the mailing list should be sent to
majordomo@csli.stanford.edu. With the lines in the body of the text
of either
subscribe csli-calendar
for the long form or
subscribe csli-short-calendar
for the short form (i.e., no abstracts). Problems with subscribing or
unsubscribing should be sent to
owner-csli-calendar@csli.stanford.edu.
The full current issue is at
http://www-csli.stanford.edu/Archive/calendar/current.shtml
and the archives at
http://www-csli.stanford.edu/Archive/calendar/
People on most of the CSLI computers can type 'help csli-calendar' to
see the current issue.
The CSLI Calendar is also posted each week to
news://news.stanford.edu/su.events
Information about CSLI's research program is available at
http://www-csli.stanford.edu/
For maps to the Stanford University campus see
http://www.stanford.edu/home/visitors/maps.html
____________