From macglashan at tti-c.org Mon May 4 15:13:14 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Mon May 4 15:13:29 2009 Subject: [TTIC Colloquium] TTI-C Talk: Jacob Eisenstein, UIUC References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: <19B6C83480814DCABA5926ACB4128860@jmacglDPLFYD1> > When: Wednesday, May 6 @ 11:00am (lunch will be provided after > talk) > > Where: 6045 S Kenwood Ave, TTI-C Conference Room #526 (5th > Floor) > > Who: Jacob Eisenstein, UIUC > > Title: Multimedia Discourse Processing > > Classroom lectures, business presentations, and other forms of face-to-face communication are increasingly archived and disseminated as multimedia. But this data is treated as a black box, and the underlying content is opaque to search and indexing technology. My research focuses on identifying high-level linguistic organization in multimedia, by leveraging visual communicative modalities. This project poses difficult challenges for both natural language processing and computer vision: visual communication is loosely-structured, idiosyncratic, and difficult to represent, and there is little labeled data. I'll describe a general approach to addressing these issues, starting from the premise that the only way forward is to reason about language and vision jointly. This approach is applied to three multimedia discourse processing tasks: topic segmentation, summarization, and resolution of noun phrase ambiguity. > Contact: Karen Livescu, TTI-C klivescu@tti-c.org > 834-2549 > > > > > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090504/b285b739/attachment-0001.htm From macglashan at tti-c.org Thu May 7 16:29:41 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Thu May 7 16:29:54 2009 Subject: [TTIC Colloquium] TTI-C Talk: Alexandru Niculescu-Mizil (IBM) References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: <2F1B49A57948456993BC067B34A29F16@jmacglDPLFYD1> > When: Thursday, May 14 @ 11:00am (lunch will be provided after > talk) > > Where: 6045 S Kenwood Ave, TTI-C Conference Room #526 (5th > Floor) > > Who: Alexandru Niculescu-Mizil (IBM T.J. Watson Research > Center) > > Title: Inductive Transfer for Bayesian Network Structure > Learning > > Bayesian Networks provide a compact, intuitive description of the dependency structure of a domain by using a directed acyclic graph to encode statistical dependencies between variables. One of the most useful features of Bayesian Networks is the ability to learn this dependency graph from observational data, and use is as a powerful data analysis tool to gain valuable insights into the problem at hand. While receiving significant attention in the machine learning community, Bayesian Network structure learning remains challenging, especially when training data is scarce. In this talk I show how structure learning performance can be significantly improved through inductive transfer, when data is available for multiple related problems. Departing from the traditional approach of learning the dependency graph for a single problem in isolation, I present a score and search algorithm for jointly learning multiple related Bayesian Networks that improves the quality of the leaned dependency structures by transferring useful information among the different related problems. I demonstrate the effectiveness of the algorithm using two standard benchmark structure learning problems, and a real bird ecology problem. > Contact: Nati Srebro, TTI-C nati@tti-c.org 834-7493 > > > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090507/dc4dca8f/attachment.htm From macglashan at tti-c.org Fri May 8 10:34:04 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Fri May 8 10:34:11 2009 Subject: [TTIC Colloquium] TTI-C Talk: Nathan Ratliff, CMU References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: <18E9419D3EA64981BDAAC462BFBE0807@jmacglDPLFYD1> > When: Monday, May 11 @ 11:00am (lunch will be provided after > talk) > > Where: 6045 S Kenwood Ave, TTI-C Conference Room #526 (5th > Floor) > > Who: Nathan Ratliff, CMU > > Title: Learning to Search: Structured Prediction Techniques > for Imitation Learning > > Modern robots successfully manipulate objects, navigate rugged terrain, drive in urban settings, and play world-class chess. Unfortunately, programming these robots is challenging, time-consuming and expensive; the parameters governing their behavior are often unintuitive, even when the desired behavior is clear and easily demonstrated. Inspired by successful end-to-end learning systems such as neural network controlled driving platforms (Pomerleau, 1989), learning-based "programming by demonstration" has gained currency as a method to achieve intelligent robot behavior. Unfortunately, with highly structured algorithms at their core, it is not clear how to effectively and efficiently train modern robotic systems using classical learning techniques. Rather than redefining robot architectures to accommodate existing learning algorithms, I develop learning techniques that leverage the performance of modern robotic components. My presentation begins with a discussion of a novel imitation learning framework we call Maximum Margin Planning which automates finding a cost function for optimal planning and control algorithms such as A*. In the linear setting, this framework has firm theoretical backing in the form of strong generalization and regret bounds. Further, I have developed practical nonlinear generalizations that are effective and efficient for real-world problems. This framework reduces imitation learning to a modern form of machine learning known as Maximum Margin Structured Classification (Taskar et al. 2005); these algorithms, therefore, apply both specifically to training existing state-of-the-art planners, as well as broadly to solving a range of structured prediction problems of importance in learning and robotics. In difficult high-dimensional planning domains, such as those found in many manipulation problems, high-performance planning technology remains a topic of much research. I will present some recent work which moves toward simultaneously advancing this technology while retaining the learnability developed above. I'll demonstrate our algorithms on a range of applications including overhead navigation, quadrupedal locomotion, heuristic learning, manipulation planning, grasp prediction, driver prediction, pedestrian prediction, optical character recognition, and LADAR classification. > Contact: Greg Shakhnarovich, TTI-C greg@tti-c.org > 834-2572 > > > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090508/4828a139/attachment.htm From cnovak at tti-c.org Mon May 11 10:24:00 2009 From: cnovak at tti-c.org (Christina Novak) Date: Mon May 11 10:27:09 2009 Subject: [TTIC Colloquium] TTI-C Distinguished Lecture Series (May 13, Dr. Steve Young- University of Cambridge) Message-ID: <481F7DE4522B430BA8F8434297EE6B3A@cnovakHBRQFD1> Good morning, Please join us for the final talk in the Toyota Technological Institute at Chicago 2009 Distinguished Lecture Series with speaker Dr. Steve Young. The talk will be held from 2 pm at TTI-C's new facility at 6045 S. Kenwood Ave., (corner of Kenwood and 61st St.). More 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. Wednesday, May 13 2pm TTI-C Conference Room 526-530 Steve Young (University of Cambridge) "Statistical Spoken Dialogue Systems" Abstract: Current spoken dialogue systems (SDS) typically employ hand-crafted decision networks or flow-charts to determine what action to take at each point in a conversation. The result is a system which is fragile to speech recognition errors and which is unable to adapt and learn from experience. Modeling a dialogue as a statistical Markov Decision Process potentially offers a way forwards. However, attempts to exploit MDPs in real systems have met with limited success primarily due to the fact that they cannot model the uncertainty which is inherent in all speech-based interactions. For the last few years, a team in the Cambridge Speech Group has been investigating the use of partially observable Markov Decision Processes (POMDPs) for use in SDS. POMDPs provide a Bayesian model of belief and a principled mathematical framework for modeling uncertainty. They can be trained from real data and they yield policies which can be optimised using reinforcement learning. However, exact belief update and policy optimisation algorithms are intractable and as a result there are many issues inherent in scaling POMDP-based systems to handle real-world tasks. This talk will briefly summarise the basic mathematics of POMDPs in SDS and explain why exact optimisation is intractable. It will then outline some of the techniques which have been developed to enable real systems to be built. The talk will conclude by presenting some working systems and results from user trials. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090511/bffccf84/attachment.htm From cnovak at tti-c.org Wed May 13 09:28:35 2009 From: cnovak at tti-c.org (Christina Novak) Date: Wed May 13 09:37:34 2009 Subject: [TTIC Colloquium] Today- Dr. Steve Young (University of Cambridge) 2-3:15pm Distinguished Lecture Series Message-ID: <11DCE417988B4F0A90C38053159391F7@cnovakHBRQFD1> Good morning, We hope to see you at the Distinguished Lecture Series talk 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. www.tti-c.org/dls Today 2pm TTI-C Conference Room 526-530 Steve Young (University of Cambridge) "Statistical Spoken Dialogue Systems" Abstract: Current spoken dialogue systems (SDS) typically employ hand-crafted decision networks or flow-charts to determine what action to take at each point in a conversation. The result is a system which is fragile to speech recognition errors and which is unable to adapt and learn from experience. Modeling a dialogue as a statistical Markov Decision Process potentially offers a way forwards. However, attempts to exploit MDPs in real systems have met with limited success primarily due to the fact that they cannot model the uncertainty which is inherent in all speech-based interactions. For the last few years, a team in the Cambridge Speech Group has been investigating the use of partially observable Markov Decision Processes (POMDPs) for use in SDS. POMDPs provide a Bayesian model of belief and a principled mathematical framework for modeling uncertainty. They can be trained from real data and they yield policies which can be optimised using reinforcement learning. However, exact belief update and policy optimisation algorithms are intractable and as a result there are many issues inherent in scaling POMDP-based systems to handle real-world tasks. This talk will briefly summarise the basic mathematics of POMDPs in SDS and explain why exact optimisation is intractable. It will then outline some of the techniques which have been developed to enable real systems to be built. The talk will conclude by presenting some working systems and results from user trials. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090513/0e8ebcd6/attachment-0001.htm From macglashan at tti-c.org Wed May 13 15:39:00 2009 From: macglashan at tti-c.org (Julia MacGlashan) Date: Wed May 13 15:39:05 2009 Subject: [TTIC Colloquium] TTI-C Talk: Alexandru Niculescu-Mizil (IBM) References: <9BFA4FFE1ACE407581765CA2326E1547@jmacglDPLFYD1> Message-ID: REMINDER When: Thursday, May 14 @ 11:00am (lunch will be provided after talk) Where: 6045 S Kenwood Ave, TTI-C Conference Room #526 (5th Floor) Who: Alexandru Niculescu-Mizil (IBM T.J. Watson Research Center) Title: Inductive Transfer for Bayesian Network Structure Learning Bayesian Networks provide a compact, intuitive description of the dependency structure of a domain by using a directed acyclic graph to encode statistical dependencies between variables. One of the most useful features of Bayesian Networks is the ability to learn this dependency graph from observational data, and use is as a powerful data analysis tool to gain valuable insights into the problem at hand. While receiving significant attention in the machine learning community, Bayesian Network structure learning remains challenging, especially when training data is scarce. In this talk I show how structure learning performance can be significantly improved through inductive transfer, when data is available for multiple related problems. Departing from the traditional approach of learning the dependency graph for a single problem in isolation, I present a score and search algorithm for jointly learning multiple related Bayesian Networks that improves the quality of the leaned dependency structures by transferring useful information among the different related problems. I demonstrate the effectiveness of the algorithm using two standard benchmark structure learning problems, and a real bird ecology problem. Contact: Nati Srebro, TTI-C nati@tti-c.org 834-7493 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://ttic.uchicago.edu/pipermail/colloquium/attachments/20090513/7e8313f5/attachment.htm