From macglashan at tti-c.org Mon Jan 5 09:24:09 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Mon Jan 5 09:21:16 2009 Subject: [TTIC Colloquium] TTI-C Talk: Ofer Neiman, Hebrew University of Jerusalem References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: <135D23B070D14556A1BC549782D61015@jmacglDPLFYD1> When: Thursday, January 8 @ 10:00am Where: TTI-C Conference Room: 1427 E. 60th St, 2nd Floor Who: Ofer Neiman, University of Jerusalem Title: Nearly Tight Low Stretch Spanning Trees We prove that any graph G on n vertices has a distribution over its spanning trees such that for any edge (u,v) the expected stretch E_T[d_T(u,v)] is bounded by \tilde{O}(\log n). Our result is obtained via a new approach of building ``highways'' between portals and a new strong diameter probabilistic decomposition theorem. Joint work with Ittai Abraham and Yair Bartal Contact: Julia Chuzhoy, TTI-C cjulia@tti-c.org 834-2490 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090105/8dc50730/attachment.htm From macglashan at tti-c.org Tue Jan 6 08:48:24 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Tue Jan 6 08:45:35 2009 Subject: [TTIC Colloquium] TTI-C Colloquium: Lance Fortnow, Northwestern University References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: <532758F2C2534AFA94304C8AB1529EF2@jmacglDPLFYD1> When: Monday, January 12 @ 2:00pm Where: TTI-C Conference Room: 1427 E. 60th St, 2nd Floor Who: Lance Fortnow, Northwestern University Title: Computational Awareness What are we aware of? More than a philosophical pursuit, awareness plays a crucial role in decision making as one cannot make a choice that one is not aware of. Billions of advertising dollars are spent in increasing the awareness of their brands. Most work on awareness has focused on logical/axiomatic approaches (for example you are aware of something if you know it or you know you don't know it). Instead we put awareness through a computational lens, roughly defining the unawareness of an object as the amount of time it takes to enumerate that object given the current environment. For example, if you want to buy a car, the ones you are most aware of are those which come earlier if you attempted to start enumerating cars. Based on Levin's "age" function, we give a formal definition of unawareness using this intuition. The formal definition uses a universal enumeration as good as any other computable enumeration up to constant factors. Our definition differs from earlier approaches as we talk about the awareness of strings as opposed to the truth of some statement and gives a quantitative measure of unawareness instead of just a binary awareness/unawareness choice. We will give some observations on how newer technologies, like search engines, have affected our awareness of various information, and how we can become unaware of objects we were once aware of. We discuss a couple of applications: - Why do loopholes occur in laws and contracts? We give an explanation based on the lack of awareness of the legislators of future circumstances and of the judge's unawareness of what the legislature's awareness. - We give a new view of sponsored search auctions based on awareness and show new bidding strategies when the advertisers wish to increase awareness of their products. This talk will cover mostly very preliminary research parts of which represent work with Kim-Sau Chung and Nikhil Devanur. Contact: Ronen Basri, TTI-C ronen.basri@tti-c.org 834-2515 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090106/9190ccb7/attachment.htm From cnovak at tti-c.org Tue Jan 6 14:42:12 2009 From: cnovak at tti-c.org (Christina Novak) Date: Tue Jan 6 14:41:03 2009 Subject: [TTIC Colloquium] TTI-C Distinguished Lecture Series JAN 26- Geoffrey Hinton (University of Toronto) Message-ID: Good afternoon, The Toyota Technological Institute at Chicago would like to invite you to our 2009 Distinguished Lecture Series on the University of Chicago campus. Our first speaker is Dr. Geoffrey Hinton. (More details on all speakers- see http://tti-c.org/dls) Monday, January 26th, 2009 Geoffrey Hinton (University of Toronto) "Recent Developments in Learning Deep Networks" Abstract: I will start by describing an efficient, modular, unsupervised learning procedure for deep generative models that contain millions of parameters and many layers of hidden features. The features are learned one layer at a time without any information about the final goal of the system. This approach leads to excellent generative models of handwritten digits. I will then describe three recent improvements. First, I will describe a better learning algorithm for the module that is used to learn each layer of features greedily. Then I will describe a more powerful type of generative module that contains multiplicative interactions so that hidden units at one level can switch in pairwise interactions between hidden units at the level below. Finally I will describe an application to recognizing stereo images of 3-D objects from the NORB database. For this task, deep belief nets outperform the best published results. All lectures will be held at 2:00pm TTI-C (Main Conference Room) 6045 S. Kenwood Avenue (5th floor) Chicago, IL 60637 We look forward to seeing you at the lecture! Sincerely, David McAllester TTI-C Chief Academic Officer For questions about the TTI-C Distinguished Lecture Series, please contact: Chrissy Novak cnovak@tti-c.org or (773)834-2216. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090106/f5f4ad7b/attachment-0001.htm From macglashan at tti-c.org Tue Jan 13 10:30:17 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Tue Jan 13 10:27:05 2009 Subject: [TTIC Colloquium] TTI-C Talk: Anup Rao, Institute for Advanced Study References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: When: Thursday, January 15 @ 10:00am Where: TTI-C Conference Room: 1427 E. 60th St, 2nd Floor Who: Anup Rao, Institute for Advanced Study Title: Hardness Amplification by Repetition We shall explore a couple of questions of the following type: If computing some functionality requires C computational resources in some model, how much resources does it take to compute k independent copies of the same function? If every machine with bounded resources can succeed in computing a functionality with probability at most 90%, what can we say about the probability of success in computing k copies of the functionality? In this talk, we shall discuss questions of this type in the context of communication protocols and PCPs and show how this study is connected to questions in geometry and hardness of approximation. Contact: Julia Chuzhoy, TTI-C cjulia@tti-c.org 834-2490 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090113/a8e024b4/attachment.htm From cnovak at tti-c.org Tue Jan 20 09:14:03 2009 From: cnovak at tti-c.org (Christina Novak) Date: Tue Jan 20 09:19:49 2009 Subject: [TTIC Colloquium] TTI-C Distinguished Lecture Series JAN 26- Geoffrey Hinton (University of Toronto) Message-ID: <812767C770D34274B05866AF834B500B@cnovakHBRQFD1> Good morning, Please join us for the Toyota Technological Institute at Chicago's first 2009 Distinguished Lecture Series speaker, Dr. Geoffrey Hinton. Monday, January 26th, 2009 (1-3pm) Geoffrey Hinton (University of Toronto) "Recent Developments in Learning Deep Networks" Abstract: I will start by describing an efficient, modular, unsupervised learning procedure for deep generative models that contain millions of parameters and many layers of hidden features. The features are learned one layer at a time without any information about the final goal of the system. This approach leads to excellent generative models of handwritten digits. I will then describe three recent improvements. First, I will describe a better learning algorithm for the module that is used to learn each layer of features greedily. Then I will describe a more powerful type of generative module that contains multiplicative interactions so that hidden units at one level can switch in pairwise interactions between hidden units at the level below. Finally I will describe an application to recognizing stereo images of 3-D objects from the NORB database. For this task, deep belief nets outperform the best published results. (More details on all speakers- see http://tti-c.org/dls) UPDATE: Lecture will be held from 1:00- 3:00pm UPDATE: Location is at the University of Chicago International House (Assembly Hall) 1414 E. 59th St. Chicago, IL 60637 We look forward to seeing you at the lecture! Sincerely, David McAllester TTI-C Chief Academic Officer For questions about the TTI-C Distinguished Lecture Series, please contact: Chrissy Novak cnovak@tti-c.org or (773)834-2216. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090120/a62ef194/attachment.htm From cnovak at tti-c.org Thu Jan 22 09:39:19 2009 From: cnovak at tti-c.org (Christina Novak) Date: Thu Jan 22 09:37:05 2009 Subject: [TTIC Colloquium] TTI-C Distinguished Lecture Series (Please note time change) Message-ID: Good morning, Please note the time change for Monday January 26th Distinguished Lecture Series talk. The talk will be from 1-3pm. Details, speaker bios and directions may be found at http://tti-c.org/dls. We look forward to seeing you at the lecture! Sincerely, David McAllester TTI-C Chief Academic Officer For questions about the TTI-C Distinguished Lecture Series, please contact: Chrissy Novak cnovak@tti-c.org or (773)834-2216. Monday, January 26th, 2009 (1-3pm) Geoffrey Hinton (University of Toronto) "Recent Developments in Learning Deep Networks" Abstract: I will start by describing an efficient, modular, unsupervised learning procedure for deep generative models that contain millions of parameters and many layers of hidden features. The features are learned one layer at a time without any information about the final goal of the system. This approach leads to excellent generative models of handwritten digits. I will then describe three recent improvements. First, I will describe a better learning algorithm for the module that is used to learn each layer of features greedily. Then I will describe a more powerful type of generative module that contains multiplicative interactions so that hidden units at one level can switch in pairwise interactions between hidden units at the level below. Finally I will describe an application to recognizing stereo images of 3-D objects from the NORB database. For this task, deep belief nets outperform the best published results. Lecture will be held from 1:00- 3:00pm Location is at the University of Chicago International House (Assembly Hall) 1414 E. 59th St. Chicago, IL 60637 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090122/0cbed517/attachment-0001.htm From cnovak at tti-c.org Mon Jan 26 08:57:00 2009 From: cnovak at tti-c.org (Christina Novak) Date: Mon Jan 26 09:04:55 2009 Subject: [TTIC Colloquium] Today- Geoffrey Hinton (Univ. of Toronto) 1-3pm Distinguished Lecture Series Message-ID: <1A80C3C2BA8044999511A2988CF57581@cnovakHBRQFD1> Good morning, We hope to see you at the lecture this afternoon! Sincerely, David McAllester TTI-C Chief Academic Officer For questions about the TTI-C Distinguished Lecture Series, please contact: Chrissy Novak cnovak@tti-c.org or (773)834-2216. Monday, January 26th, 2009 (1-3pm) Geoffrey Hinton (University of Toronto) "Recent Developments in Learning Deep Networks" Abstract: I will start by describing an efficient, modular, unsupervised learning procedure for deep generative models that contain millions of parameters and many layers of hidden features. The features are learned one layer at a time without any information about the final goal of the system. This approach leads to excellent generative models of handwritten digits. I will then describe three recent improvements. First, I will describe a better learning algorithm for the module that is used to learn each layer of features greedily. Then I will describe a more powerful type of generative module that contains multiplicative interactions so that hidden units at one level can switch in pairwise interactions between hidden units at the level below. Finally I will describe an application to recognizing stereo images of 3-D objects from the NORB database. For this task, deep belief nets outperform the best published results. Lecture will be held from 1:00- 3:00pm Location is at the University of Chicago International House (Assembly Hall) 1414 E. 59th St. Chicago, IL 60637 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090126/4bdad1d1/attachment.htm From macglashan at tti-c.org Mon Jan 26 09:23:06 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Mon Jan 26 09:19:11 2009 Subject: [TTIC Colloquium] TTI-C Talk: Alexander Sherstov (University of Texas at Austin) Message-ID: <7CA87722A5AA44AA914B66C4EB2B593A@jmacglDPLFYD1> *** TTI-C is moving! This talk will be held at our new location, one block southwest of the Press Building. You can see a map here: http://maps.uchicago.edu/farsoutheast/ *** When: Monday, February 2 @ 11:00am Where: TTI-C Conference Room: 6045 S Kenwood Ave, 5th Floor Who: Alexander Sherstov (University of Texas at Austin) Title: Lower Bounds for Communication Complexity Using Pattern Matrices Communication complexity studies the amount of communication necessary to compute a given function when its arguments are distributed among several parties. This research area plays a fundamental role in theoretical computer science and beyond, with applications as diverse as circuit complexity, learning theory, pseudorandomness, and private information retrieval. I will present the pattern matrix method, a new approach that I have developed for studying communication complexity, based on analytic techniques such as linear-programming duality and approximation theory. Using my method, I will show how to solve several longstanding problems in the field. As a first such application, I will resolve a well-known question in circuit complexity on the limitations of neural networks. In particular, I will prove the optimality of Allender's classic simulation of AND/OR/NOT circuits by neural networks. Second, I will show that polynomial-size formulas in disjunctive normal form (DNF) have a complicated geometry, namely, they cannot be embedded in halfspaces of subexponential dimension. This solves an open problem in communication complexity and helps explain the lack of progress on the computational learning of DNF formulas. As a final application, I will prove that for a large and natural class of problems, quantum communication is not more powerful than classical communication. I will also discuss the broader impact of the pattern matrix method, surveying a series of important advances by other researchers. Contact: Julia Chuzhoy, TTI-C cjulia@tti-c.org 834-2490 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090126/f40dff80/attachment.htm From macglashan at tti-c.org Thu Jan 29 11:10:12 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Thu Jan 29 11:09:28 2009 Subject: [TTIC Colloquium] TTI-C Talk: Xi Chen (Princeton University) Message-ID: <96B8E4CA44BB43968A3D8E24A9D2177E@jmacglDPLFYD1> *** TTI-C moved! This talk will be held at our new location, one block southwest of the Press Building. You can see a map here: http://maps.uchicago.edu/farsoutheast/ *** When: Tuesday, February 3 @ 11:00am Where: TTI-C Conference Room: 6045 S Kenwood Ave, 5th Floor Who: Xi Chen (Princeton University) Title: On the Computation and Approximation of Two-Player Nash Equilibria In 1950, Nash showed that every non-cooperative game has an equilibrium. Before his work, the result was known only for two-player zero-sum games. While von Neumann's minimax theorem was the mathematical foundation of the two-player zero-sum game, Brouwer's and Katutani's fixed point theorems were crucially used in Nash's proofs. His approach was later used by Arrow and Debreu in their development of the General Equilibrium Theory. Fourteen years after Nash's work, Lemke and Howson designed a simplex-like path-following algorithm for finding a Nash equilibrium in a general two-player game. Despite its several appealing properties, this algorithm was recently shown to have exponential worst-case complexity. In contrast, in his groundbreaking work, Khachiyan showed that the ellipsoid algorithm can solve any linear program and hence any two-player zero-sum game in polynomial time. However, this success has not been extended to general two-player games --- no polynomial-time algorithm has yet been found for this remarkable problem. In this talk, I will present results that characterize the complexity of computing and approximating two-player Nash equilibria (Joint work with Xiaotie Deng and Shang-Hua Teng). Contact: Lance Fortnow fortnow@eecs.northwestern.edu -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090129/45281671/attachment-0001.htm From macglashan at tti-c.org Thu Jan 29 11:18:20 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Thu Jan 29 11:17:44 2009 Subject: [TTIC Colloquium] TTI-C Colloquium: Matthew Stephens, University of Chicago References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: <7A93567F4F3B45A198BF4B8BD681B77E@jmacglDPLFYD1> *** TTI-C moved! This talk will be held at our new location, one block southwest of the Press Building. You can see a map here: http://maps.uchicago.edu/farsoutheast/ *** When: Monday, February 2 @ 2:00pm Where: TTI-C Conference Room #526, 6045 S Kenwood Ave, 5th Floor Who: Matthew Stephens, University of Chicago Title: Large-scale Bayesian regression: A novel prior distribution, with applications to genome wide association studies Ongoing large-scale genetic association studies, in an attempt to identify variants and genes affecting susceptibility to common diseases, are typing hundreds of thousands of genetic variants (SNPs) in thousands of individuals, with the aim of identifying which SNPs are associated with disease status. Standard statistical analyses consider each SNP in turn, and perform hundreds of thousands of univariate tests. Here we consider analysing all SNPs simultaneously using Bayesian variable-selection multiple regression (in the context of a continuous response). We suggest a novel prior distribution for the effect sizes of associated SNPs, based on the overall proportion of variance explained. We assess the computational feasibility of this approach through simulation, and investigate its merits relative to standard univariate analyses. Our methods should be of general interest to those interested in sparse regression and variable selection problems. Contact: David McAllester, TTI-C mcallester@tti-c.org -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090129/856a8924/attachment.htm From macglashan at tti-c.org Thu Jan 29 11:32:57 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Thu Jan 29 11:32:22 2009 Subject: [TTIC Colloquium] TTI-C Talk: Prasad Raghavendra (University of Washington) References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: *** TTI-C moved! This talk will be held at our new location, one block southwest of the Press Building. You can see a map here: http://maps.uchicago.edu/farsoutheast/ *** When: Wednesday, February 4 @ 11:00am Where: TTI-C Conference Room #526, 6045 S Kenwood Ave, 5th Floor Who: Prasad Raghavendra (University of Washington) Title: Approximating the optimum: Efficient algorithms and their limits Most combinatorial optimization problems of interest are NP-hard to solve exactly. To cope with this intractability, one settles for approximation algorithms with provable guarantee on the quality of approximation. Despite great success in designing approximation algorithms, underlying a vast majority of the work is the technique of linear programming, or more generally semi-definite programming. This poses the natural question: How far can we push the technique of semi-definite programming? Are there techniques that yield better approximations than semi-definite programming? In this work, we show that the simplest semi-definite programs yield the best approximation for large classes of optimization problems ranging from constraint satisfaction problems (CSP) to graph labeling problems, ordering CSPs to certain clustering problems, under the Unique Games Conjecture. In a surprising convergence of algorithms and hardness results, the same work also leads to a generic algorithm that achieves the optimal approximation for every constraint satisfaction problem. Bio: Prasad Raghavendra obtained his bachelor's degree at Indian Institute of Technology, Madras, and is currently a Phd candidate at University of Washington. His research spans several areas of theoretical computer science including approximation algorithms and inapproximability results for NP-hard optimization problems, error correcting codes, metric embeddings, computational learning and complexity. He is the recipient of the Best Paper and Best Student Paper awards at STOC 08. Contact: Julia Chuzhoy, TTI-C cjulia@tti-c.org 834-2490 -------------- next part -------------- A non-text attachment was scrubbed... Name: winmail.dat Type: application/ms-tnef Size: 8626 bytes Desc: not available Url : http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090129/147c5873/winmail.bin From macglashan at tti-c.org Fri Jan 30 12:43:16 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Fri Jan 30 12:42:47 2009 Subject: [TTIC Colloquium] TTI-C Talk: Yury Makarychev (Microsoft Research) References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: <717214A4750E4524B7BB556B054EEAA9@jmacglDPLFYD1> *** TTI-C moved! This talk will be held at our new location, one block southwest of the Press Building. You can see a map here: http://maps.uchicago.edu/farsoutheast/ *** When: Thursday, February 5 @ 11:00am Where: TTI-C Conference Room #526, 6045 S Kenwood Ave, 5th Floor Who: Yury Makarychev (Microsoft Research) Title: Approximation Algorithms, Semidefinite Programming and Unique Games Best known approximation algorithms for many combinatorial optimization problems are based on semidefinite (SDP) programming. Are these algorithms optimal? Are there any techniques better than SDP? We do not know the answer. However, the Unique Games Conjecture (UGC) implies that for a wide class of problems known SDP algorithms are indeed optimal. One of the central questions in the theory of approximation algorithms is whether this conjecture is true or false. To answer this question we need to better understand Unique Games and design good approximation algorithms for them. In the first part of the talk, we will describe approximation algorithms for Unique Games that significantly improve previously known approximation guarantees and that are asymptotically optimal assuming UGC. Then we will discuss other possible approaches to Unique Games. We will describe the Sherali-Adams (SA) hierarchy of relaxations for combinatorial optimization problems (the strongest LP hierarchy considered in the literature). We will prove negative results for MAX CUT, Vertex Cover, Max Acyclic Subgraph and other problems. In particular, we will show that algorithms based on SA relaxations cannot disprove UGC. Based on joint papers with Moses Charikar, Eden Chlamtac, and Konstantin Makarychev. Contact: Julia Chuzhoy, TTI-C cjulia@tti-c.org 834-2490 -------------- next part -------------- A non-text attachment was scrubbed... Name: winmail.dat Type: application/ms-tnef Size: 8210 bytes Desc: not available Url : http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090130/fa88add9/winmail-0001.bin