Select a calendar:
Filter February Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Events for February 01, 2016
-
Online Dynamic Robust PCA
Mon, Feb 01, 2016 @ 10:00 AM - 11:00 AM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Prof. Namrata Vaswani, Electrical and Computer Engineering, Iowa State University
Talk Title: Online Dynamic Robust PCA
Abstract: We introduce a novel and provably correct solution approach, called ReProCS, to the online dynamic robust principal components' analysis (PCA) problem. Robust PCA (RPCA) can be understood as a problem of separating a low-rank matrix of the true data, L, and a sparse matrix of outliers, S, from their sum, Y = L + S. Application domains include computer vision and data analytics, among others. For example, the problem of separating sparse foregrounds (e.g., moving objects) from slowly changing backgrounds in video sequences can be posed as an instance of RPCA. This is a key first step in simplifying many computer vision tasks, e.g., video surveillance, low-bandwidth mobile video chats and video conferencing, low-light imaging ("seeing moving objects in the dark") and video denoising. RPCA solutions are also very useful in solving product recommender systems' design problems, such as the Netflix problem, when the user data may contain outliers (e.g., due to lazy or malicious users). While there has been a large amount of recent work on provably correct batch RPCA solutions, the online and dynamic RPCA problem is largely open. Online dynamic RPCA is the problem of solving RPCA on-the-fly, with the extra assumptions that the initial subspace is accurately known and that the subspace from which the true data is generated is either fixed or changes slowly over time. For most of the applications discussed above, an online solution is clearly preferable and it can be argued that these extra assumptions hold. We demonstrate the power of our proposed ReProCS based online dynamic RPCA solution for many of the above applications. Moreover, under mild assumptions, we show that, with high probability, ReProCS recovers the support of the outliers exactly at all times; the subspace in which the true data lies is tracked accurately; and the error in the estimates of both is small at all times.
Host: Professor Mahdi Soltanolkotabi
Location: 248
Audiences: Everyone Is Invited
Contact: Talyia Veal
-
Communications, Networks & Systems (CommNetS) Seminar
Mon, Feb 01, 2016 @ 11:00 AM - 12:00 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Achilleas Anastasopoulos, Univ. of Michigan, Ann Arbor
Talk Title: A systematic process for evaluating structured equilibria in dynamic games with asymmetric information
Series: CommNetS
Abstract: We consider problems involving multiple agents making decisions dynamically in the presence of asymmetric information.
When agents have a common objective (dynamic decentralized teams) recent results have established a systematic framework for obtaining the optimal decision strategy that is akin to the well-known backward induction in partially observed Markov decision processes (POMDPs).
However, when agents are strategic (dynamic games with asymmetric information) there is no known systematic process for evaluating the appropriate equilibria in a sufficiently general setting. The well-known backward induction process for finding sub-game perfect equilibria is useless in these problems and we are stuck with an indecomposable fixed-point equation in the space of strategies and beliefs.
In this talk we will discuss a class of perfect Bayesian equilibria (PBE) that are the counterparts of Markov perfect equilibria (MPE) for asymmetric information games. The corresponding "state" is a belief based on the common information among agents.
We will then propose a two-step backward-forward inductive algorithm to find these structured PBE. The backward inductive part of this algorithm defines an equilibrium generating function. Each period in the backward induction involves solving a "small" fixed point equation. Using this generating function, equilibrium strategies and beliefs are defined through a forward recursion.
Biography: Achilleas Anastasopoulos received the Diploma in EE from the National Technical University of Athens, Greece in 1993, and the M.S. and Ph.D. degrees in EE from the University of Southern California in 1994 and 1999, respectively. He is currently an Associate Professor of EECS at the University of Michigan, Ann Arbor. His research interests lie in 1) the general area of communication and information theory, with emphasis in channel coding and multi-user channels; 2) control theory with emphasis in decentralized stochastic control and its connections to communications and information-theoretic problems; 3) analysis of dynamic games and mechanism design for resource allocation in networked systems.
Host: Dr. Ashutosh Nayyar
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Annie Yu
-
Seminars in Biomedical Engineering
Mon, Feb 01, 2016 @ 12:30 PM - 01:49 PM
Alfred E. Mann Department of Biomedical Engineering
Conferences, Lectures, & Seminars
Speaker: Dion Kai Dickman, PhD, Assistant Professor in Neurobiology, USC Dornsife
Talk Title: Homeostatic control of sleep and synaptic plasticity
Abstract: Homeostatic Control of Sleep and Synaptic Plasticity
Summary
Synapses have the remarkable ability to adaptively modulate synaptic strength in response to perturbations that would otherwise destabilize neurotransmission, referred to as homeostatic synaptic plasticity. Homeostatic signaling systems have emerged as robust and potent regulators of neural activity, enabling stable synaptic function while permitting the flexibility necessary for learning and memory, yet the molecules and mechanisms involved remain poorly understood. We have pioneered forward genetic approaches in Drosophila to identify genes required for homeostatic synaptic plasticity. We will first discuss an enigmatic protein complex that has emerged from this screen, which is associated with schizophrenia, and the role of this complex in synaptic function and homeostatic plasticity. We will then present data about how an individual synapse adapts to conflicting homeostatic perturbations to stable synaptic function. Finally, we are developing new tools, including translational profiling and light sheet microscopy, to reveal homeostatic adaptations to synaptic function, which may be linked to sleep, and ancient, essential, and fundamental homeostatic signaling system shared by all animal life.
Biography: Bio
Dion Dickman was born in Hawaii and did his undergraduate work at Washington University in St. Louis, studying synaptogenesis at the mouse neuromuscular junction in the lab of Joshua Sanes. He went to Harvard for graduate work and UCSF for his postdoctoral studies, performing electrophysiology-based, forward genetic screens in Drosophila, identifying new genes involved in synaptic development, function, and plasticity. He has recently started his own laboratory at the University of Southern California, where his group investigates how synaptic transmission is kept within stable physiological ranges in the nervous system, while still permitting the flexibility necessary for learning and memory. Using Drosophila as our model system, we are interested in the genes and molecular mechanisms that achieve and maintain the homeostatic control of synaptic strength, and how dysfunction in this process may contribute to neuropsychiatric disease. We are using a combination of genetic, electrophysiological, imaging, and behavioral approaches to gain insight into this complex and fundamental form of neural plasticity.
http://dornsife.usc.edu/labs/dickmanlab
Host: K. Kirk Shung, PhD
Location: Olin Hall of Engineering (OHE) - 122
Audiences: Everyone Is Invited
Contact: Mischalgrace Diasanta
-
Faculty Candidate Seminar
Mon, Feb 01, 2016 @ 01:00 PM - 02:00 PM
Daniel J. Epstein Department of Industrial and Systems Engineering
Conferences, Lectures, & Seminars
Speaker: Private, Private
Talk Title: Modeling Disease for Effective Control - Tuberculosis in India
Host: Epstein Department of ISE
Location: Ethel Percy Andrus Gerontology Center (GER) - 206
Audiences: Everyone Is Invited
Contact: Michele ISE
