Past Events

Abstract:

Learning the structure of graphical models from data is a fundamentaltask in many scientific domains. I will describe analysis andapplications of learning tree-structured graphical models via awell-known learning algorithm known as the Chow-Liu algorithm (1968).The Chow-Liu algorithm learns the maximum-likelihood (ML) treestructure from independently drawn observations of a multivariatediscrete tree distribution. Using the theory of large-deviations, weanalyze the error exponent that the ML-estimate of the Markov treestructure differs from the true tree.

ABSTRACT:

Abstract: We present a brief historical account of paths leading to thepresent interest in cyberphysical systems. We follow this up withan account of several foundational research topics that underliethis area. These include issues in data fusion,real-time communication, clock synchronization, security,middleware, hybrid systems and proofs of correctness. Bio:

Abstract Consider the setting of (adversarial) online convex optimization: an adversary and a learner (or optimizer) interact in a game with T rounds. At round t, the learner has to choose a point x_t, the adversary then responds with a convex function f_t and the learner suffers f_t(x_t). The goal of the learner is to minimize its regret: a quantity that measures the "displeasure" or regret the learner feels, in hindsight, because of not knowing the future and thus behaving sub-optimally.

Abstract Cellular networks are usually modeled by placing the base stations on a grid, with mobile users either randomly scattered or placed deterministically. These models have been used extensively but suffer from being both highly idealized and not very tractable, so complex system-level simulations are used to evaluate coverage/outage probability and rate. More tractable models have long been desirable. We develop new general models for multi-cell SINR using stochastic geometry.

Abstract: We consider a sequential decision problem where at each time step the decision maker has to choose how to distribute the future loss between k alternatives, and then observes the loss of each alternative. Motivated by load balancing and job scheduling, we consider a global cost function (over the losses incurred by each alternative), rather than a summation of the instantaneous losses as done traditionally in online learning. Such global cost functions include the makespan (the maximum over the alternatives) and the $L_d$ norm (over the alternatives).

Over the past few years there has been a rapidly growing interest in analysis, design and optimization of various types of collective behaviors in networked dynamic systems. Collective phenomena (such as flocking, schooling, rendezvous, synchronization, and agreement) have been studied in a diverse set of disciplines, ranging from computer graphics and statistical physics to distributed computation, and from robotics and control theory to social sciences and economics.

Abstract:

Abstract: Wireless protocol design today is based on a simple model of howwireless channel behaves. This model does not capture the temporal andspatial correlations of packet reception that wireless linksexperience in reality. Unfortunately, moving beyond this model hasbeen difficult due to the lack of good ways to quantify wireless linkcomplexities.