Navigation

Tutorials

Tutorial 1: Simultaneous perturbation methods for stochastic non-convex optimization
Prashanth L. A., Assistant Professor in the Department of Computer Science and Engineering at Indian Institute of Technology Madras

Abstract:
Stochastic recursive algorithms form the basis of several optimization approaches and find wide applicability across various engineering disciplines such as machine learning, communication engineering, signal processing and robotics. The highly efficient simultaneous perturbation approaches have been considered as the unifying thread in all these algorithms. This tutorial presents algorithms for smooth non-convex optimization based on the simultaneous perturbation method. We may mention here that in this proposal, by simultaneous perturbation methods, we refer to the entire family of algorithms that are based on either gradient or gradient and Hessian estimates that are obtained using some form of simultaneous random perturbations. A remarkable feature of the algorithms is that they are easily implementable, do not require an explicit system model, and work with real or simulated data. The tutorial also covers applications in sensor networks and service systems to illustrate these points.

Bio:
Prashanth L.A. is an Assistant Professor in the Department of Computer Science and Engineering at Indian Institute of Technology Madras. Prior to this, he was a postdoctoral researcher at the Institute for Systems Research, University of Maryland - College Park from 2015 to 2017 and at INRIA Lille - Team SequeL from 2012 to 2014. From 2002 to 2009, he was with Texas Instruments (India) Pvt Ltd, Bangalore, India. He received his Masters and Ph.D degrees in Computer Science and Automation from Indian Institute of Science, in 2008 and 2013, respectively. He was awarded the third prize for his Ph.D. dissertation, by the IEEE Intelligent Transportation Systems Society (ITSS). He is the coauthor of a book entitled `Stochastic Recursive Algorithms for Optimization: Simultaneous Perturbation Methods', published by Springer in 2013. His research interests are in reinforcement learning, stochastic optimization and multi-armed bandits, with applications in transportation systems, wireless networks and recommendation systems.

Tutorial 2: Distributed optimization in graphical models
Jinwoo Shin, Associate professor at the School of Electrical Engineering at KAIST, Korea

Abstract:
Graphical model (GM) has been one of the powerful paradigms for succinct representations of joint probability distributions in various scientific fields including machine learning, information theory, statistical physics. GM represents a joint distribution of some random variables by a graph structured model where each vertex corresponds to a random variable and each edge captures the conditional dependence between random variables. Two major algorithmic tasks arising in GM's applications are inference and learning. First, for inference, I will overview two most popular approaches: markov chain monte carlo (MCMC) and variational methods (VM). Second, for learning, I will describe the popular contrastive divergence approach which uses both MCMC and VM as its sub-routines. All inference and learning algorithms introduced in this tutorial have distributed natures and I will also present their applications to wireless and social networks.

Bio:
Jinwoo Shin is currently an associate professor at the School of Electrical Engineering at KAIST, Korea. His current major research interest is on algorithmic questions for machine learning and networking. He obtained the Ph.D. degree from Massachusetts Institute of Technology in 2010 with George M. Sprowls (Best MIT CS PhD Thesis) Award and B.S. degrees (in Math and CS) from Seoul National University in 2001. After spending two years (2010-2012) at Algorithms & Randomness Center, Georgia Institute of Technology, one year (2012-2013) at Business Analytics and Mathematical Sciences Department, IBM T. J. Watson Research, he joined KAIST EE in Fall 2013. He received Best Publication Award from INFORMS Applied Probability Society 2013 and ACM SIGMETRICS Rising Star Award 2015, in addition to best/oral papers at SIGMETRICS, NIPS and MOBIHOC. He is currently an associate editor of IEEE/ACM Transactions on Networking, ACM Modeling and Performance Evaluation of Computing Systems, and has served TPCs at AAAI, INFOCOM, INFORMS, MOBIHOC, NIPS, SIGMETRICS, WIOPT.

Schedule

9:00 am - 10:30 am: Tutorial 1, Lecture 1
10:30 am -11:00 am: Coffee break
11:00 am - 12:30 pm: Tutorial 1, Lecture 2
12:30 pm - 1:30 pm: Lunch
1:30 pm - 3:00 pm: Tutorial 2, Lecture 1
3:00 pm - 3:30 pm: Coffee break
3:30 pm - 5:00 pm: Tutorial 2, Lecture 2