Select a calendar:
Filter October Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Conferences, Lectures, & Seminars
Events for October
-
CS Colloquium: Rong Ge (Duke University) - Avoid Spurious Local Optima: Homotopy Method for Tensor PCA
Thu, Oct 06, 2016 @ 04:00 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Rong Ge, Duke University
Talk Title: Avoid Spurious Local Optima: Homotopy Method for Tensor PCA
Series: Yahoo! Labs Machine Learning Seminar Series
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium. Part of Yahoo! Labs Machine Learning Seminar Series.
Recently, several non-convex problems such as tensor decomposition, phase retrieval and matrix completion are shown to have no spurious local minima, which allows them to be solved by very simple local search algorithms. However, more complicated non-convex problems such as the Tensor PCA do have local optima that are not global, and previous results rely on techniques inspired by Sum-of-Squares hierarchy. In this work we show the commonly applied homotopy method, which tries to solve the optimization problem by considering different levels of "smoothing", can be applied to tensor PCA and achieve similar guarantees as the best known Sum-of-Squares algorithms. This is one of the first settings where local search algorithms are guaranteed to avoid spurious local optima even in high dimensions.
This is based on joint work with Yuan Deng (Duke University).
Biography: Rong Ge is an assistant professor at Duke computer science department. He got his Ph.D. in Princeton University and was a post-doc at Microsoft Research New England before joining Duke. Rong Ge is broadly interested in theoretical computer science and machine learning. His research focuses on designing algorithms with provable guarantees for machine learning problems, with applications to topic models, sparse coding and computational biology.
Host: Yan Liu
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Keshav Pingali (UT Austin) - Parallel Programming Needs Data-centric Foundations
Tue, Oct 18, 2016 @ 04:00 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Keshav Pingali , UT Austin
Talk Title: Parallel Programming Needs Data-centric Foundations
Series: CS Colloquium
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium.
Multicore and manycore processors are now ubiquitous, but
parallel programming remains as difficult as it was 30-40 years ago. During this time, our community has explored many promising approaches including functional and dataflow languages, logic programming, and automatic parallelization using program analysis and restructuring, but none of these approaches has succeeded except in a few niche application areas.
In this talk, I will argue that these problems arise largely from the computation-centric foundations and abstractions that we currently use to think about parallelism. In their place, I will propose a novel data-centric foundation for parallel programming called the operator formulation in which algorithms are described in terms of actions on data. The operator formulation shows that a generalized form of data-parallelism called amorphous data-parallelism is ubiquitous even in complex, irregular applications such as mesh generation/refinement/partitioning and SAT solvers. Regular algorithms emerge as a special case of irregular ones,
and many application-specific optimization techniques can be generalized to a broader context. The operator formulation also leads to a structural analysis of algorithms called TAO-analysis that provides implementation guidelines for exploiting parallelism efficiently. Finally, I will describe a system called Galois based on these ideas for exploiting amorphous data-parallelism on multicores and GPUs.
Biography: Keshav Pingali is a Professor in the Department of Computer Science at the University of Texas at Austin, and he holds the W.A."Tex" Moncrief Chair of Computing in the Institute for Computational Engineering and Sciences (ICES) at UT Austin. He was on the faculty of the Department of Computer Science at Cornell University from 1986 to 2006, where he held the India Chair of Computer Science.
Pingali is a Fellow of the IEEE, ACM and the AAAS. He was the co-Editor-in-chief of the ACM Transactions on Programming Languages and Systems, and currently serves on the editorial boards of the ACM Transactions on Parallel Computing, International Journal of Parallel Programming and Distributed Computing. He has also served on the NSF CISE Advisory Committee.
Host: Chao Wang
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CAIS Seminar: Stefano Ermon (Stanford) - Measuring Progress towards Sustainable Development Goals with Machine Learning
Wed, Oct 19, 2016 @ 04:00 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Stefano Ermon, Stanford University
Talk Title: Measuring Progress towards Sustainable Development Goals with Machine Learning
Series: Center for AI in Society (CAIS) Seminar Series
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium.
Recent technological developments are creating new spatio-temporal data streams that contain a wealth of information relevant to sustainable development goals. Modern AI techniques have the potential to yield accurate, inexpensive, and highly scalable models to inform research and policy. As a first example, I will present a machine learning method we developed to predict and map poverty in developing countries. Our method can reliably predict economic well-being using only high-resolution satellite imagery. Because images are passively collected in every corner of the world, our method can provide timely and accurate measurements in a very scalable end economic way, and could revolutionize efforts towards global poverty eradication. As a second example, I will present some ongoing work on monitoring agricultural and food security outcomes from space.
Biography: Stefano Ermon is currently an Assistant Professor in the Department of Computer Science at Stanford University, where he is affiliated with the Artificial Intelligence Laboratory and the Woods Institute for the Environment. He completed his PhD in computer science at Cornell in 2015. His research interests include techniques for scalable and accurate inference in graphical models, statistical modeling of data, large-scale combinatorial optimization, and robust decision making under uncertainty, and is motivated by a range of applications, in particular ones in the emerging field of computational sustainability. Stefano has won several awards, including two Best Student Paper Awards, one Runner-Up Prize, and a McMullen Fellowship.
Host: Milind Tambe
Location: Mark Taper Hall Of Humanities (THH) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium and RASC seminar: Ali Agha (Caltech, JPL) - Quantifiably safe robot motion planning under motion and sensing uncertainty
Thu, Oct 20, 2016 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Ali Agha, Caltech, JPL
Talk Title: Quantifiably safe robot motion planning under motion and sensing uncertainty
Series: RASC Seminar Series
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium.
Planning robot motions amidst obstacles, while actively enhancing localization, is a key component for true autonomy. With a growing number of autonomous robots and safety-critical applications, it is of paramount importance to design planners with the ability to guarantee and quantify the system's safety. In this talk, we explore planning methods that reason about the acquisition of future perceptual knowledge and incorporate this knowledge in planning to accurately quantify the success probability and safety of the plan. In particular, I present a planning framework under motion and sensing uncertainty, called Feedback-based Information RoadMap (FIRM). FIRM is a multi-query graph in belief space (space of probability distributions), which can be viewed as the belief space variant of the celebrated PRM (probabilistic roadmap). Each node of FIRM is a belief. Each edge (belief-to-belief transition) is realized via composition of closed-loop controllers that behave like funnels in belief space. We also discuss the feedback nature and scalability of the generated plan. We will demonstrate this approach in the context of robot navigation in indoor GPS-denied environments.
Biography: Ali-Akbar Agha-Mohammadi is a Robotics Research Technologist at NASA JPL/California Institute of Technology. Previously, he was a research engineer at Qualcomm Research and a post-doctoral researcher at LIDS/ACL at MIT. He has received his Ph.D. in Computer Science and Engineering from Texas A&M. He also holds B.S. and M.S. degrees in Electrical Engineering (Control Systems). His research interests include robotics, stochastic systems, control systems, estimation, and filtering theory.
Host: CS Department
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS PhD Colloquium: Om Prasad Patri (USC) - Shape Mining for Multisensor Event Recognition
Tue, Oct 25, 2016 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Om Prasad Patri, USC
Talk Title: Shape Mining for Multisensor Event Recognition
Series: CS Colloquium
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium.
CS PhD Colloquium Lecture Event.
The ubiquitous rise of sensors in our daily lives, as well in industrial and engineering equipment, have led to emerging challenges in pattern analysis of large amounts of multisensor data to identify critical events automatically. This talk presents our recent work on framing this event recognition problem in the context of time series classification by automatically finding discriminative shapes or patterns (called shapelets) within sensor data. These unconventional shape mining approaches show potential for real-world sensor datasets, such as equipment monitoring data from an oil field or a manufacturing plant, as they don't make assumptions about the nature and structure of the input sensor data and provide visual intuition in the form of the extracted shapes for further analysis by domain experts, instead of being a black-box machine learning approach.
These approaches also perform fast classification as they focus on throwing away most of the data after finding the discriminative shapelets. By combining shape extraction and feature selection, this temporal pattern mining paradigm can be extended for processing data from multiple sensors. This talk describes algorithmic strategies for performing this combination, and presents results and motivational examples from modern industrial systems where our approaches are applicable. An interesting application of the proposed approach using shape mining for identifying malware from Windows portable executable files is also discussed.
Biography: Om P. Patri is a PhD candidate in Computer Science at USC, advised by Prof. Viktor K. Prasanna. His interests are broadly in the areas of data science, AI, cybersecurity and event-based systems, and his dissertation research is on modeling and recognition of events from multisensor time series data. At USC, he has been a part of the Center for Smart Interactive Oil Field Technologies (CiSoft) and the USC Data Science Lab. During his graduate studies at USC, he has worked with NEC Labs America and Cylance Inc. Before coming to USC, he obtained a Bachelors in Computer Science from the Indian Institute of Technology Guwahati.
Host: CS Department
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Sewoong Oh (UIUC) - Fundamental Limits and Efficient Algorithms in Adaptive Crowdsourcing
Thu, Oct 27, 2016 @ 04:00 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Sewoong Oh , UIUC
Talk Title: Fundamental Limits and Efficient Algorithms in Adaptive Crowdsourcing
Series: Yahoo! Labs Machine Learning Seminar Series
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium. Part of Yahoo! Labs Machine Learning Seminar Series.
Adaptive schemes, where tasks are assigned based on the data collected thus far, are widely used in practical crowdsourcing systems to efficiently allocate the budget. However, existing theoretical analyses of crowdsourcing systems suggest that the gain of adaptive task assignments is minimal. To bridge this gap, we propose a new model for representing practical crowdsourcing systems, which strictly generalizes the popular Dawid-Skene model, and characterize the fundamental trade-off between budget and accuracy. We introduce a novel adaptive scheme that matches this fundamental limit. We introduce new techniques to analyze the spectral analyses of non-back-tracking operators, using density evolution techniques from coding theory.
Biography: Sewoong Oh is an Assistant Professor of Industrial and Enterprise Systems Engineering at UIUC. He received his PhD from the department of Electrical Engineering at Stanford University. Following his PhD, he worked as a postdoctoral researcher at Laboratory for Information and Decision Systems (LIDS) at MIT. He was co-awarded the Kenneth C. Sevcik outstanding student paper award at the Sigmetrics 2010, the best paper award at the SIGMETRICS 2015, and NSF CAREER award in 2016.
Host: Yan Liu
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
