[Ml-stat-talks] Blackboard Sessions at Princeton: Dimitris Achliotas on Friday March 21st.

Philippe Rigollet rigollet at Princeton.EDU
Tue Mar 11 11:18:08 EDT 2014


Hi Moses, 

I’ve just sent it to ml-stat and theory-read but if you know other lists who may be interested (IAS,…), please feel free to circulate.
Thanks
Philippe
—————

I am very pleased to announce the first Blackboard sessions at Princeton.

The purpose of the sessions is to give an introduction to/overview of interesting ideas, results, techniques, problems, and challenges in an exciting area of research.  This is a fantastic opportunity for our faculty and students to be introduced to an exciting research direction, and potentially to initiate new research projects and interactions. The topics are chosen to attract a broad audience. No preliminary knowledge of the field is required. The talks are the blackboard.

The format of the sessions is:

10:30 AM - 12:00 PM: Morning session
12:00 PM - 1:30 PM:  Lunch break
1:30 PM  - 3:00 PM:  Afternoon session


Our first speaker will be Dimitris Achlioptas from UC Santa Cruz.


LOCATION: Sherrerd Hall, room 101.
DATE and TIME: Friday, March 21st. 10:30 AM - 12:00 PM and 1:30 PM  - 3:00 PM.

TITLE: Algorithmic Barriers from Phase Transitions
ABSTRACT: The study of randomly generated Constraint Satisfaction Problems (CSPs) in Computer Science began about 25 years ago. From the outset, notions from statistical physics were invoked, primarily “phase transitions”. It took computer scientists and physicists about a decade after that point to figure out that “diluted mean-field spin glasses” are “random CSPs” (and vice versa). After that rather slow start, a very fruitful and exciting exchange of ideas between the two fields has been taking place, at an accelerating pace. Central to this exchange is the heuristic notion that symmetry breaking, by introducing long-range correlations, can create computational hardness. In these sessions I will survey our current understanding of this central point, discuss open problems, and some connections to problems in communication and machine learning.  

SPEAKER BIO: Dimitris Achlioptas joined the Department of Computer Science of UC Santa Cruz in 2005 after being with Microsoft Research, Redmond from 1998. In theory, his expertise lies in the interaction between randomness and computation and his work on that topic has appeared in journals including Nature, Science, and the Annals of Mathematics. For that work he has received an NSF CAREER award, a Sloan Fellowship, and an IDEAS Starting Grant from the European Research Council. In practice, he likes to think about scalability questions and holds 18 US Patents on topics ranging from load balancing and cache optimization to web search personalization. In his free time he enjoys overworking.

--
Philippe Rigollet
www.princeton.edu/~rigollet





More information about the Ml-stat-talks mailing list