OR Seminar: Derivative-Free Nonconvex Stochastic Optimization with Application to an Energy Storage Problem with Saeed Ghadimi

Thursday, January 23, 2020

Bahen Centre, Room 1220
40 St. George Street

This event is open to the public and registration is not required.

View all upcoming Operations Research Seminars



In this talk, we propose and analyze derivative-free stochastic approximation algorithms for nonconvex optimization. We first propose generalization of the conditional gradient algorithm achieving a similar rate to the standard stochastic gradient algorithm (SGD) using only noisy function evaluations (zeroth-order information).  For the high-dimensional setting, we explore the advantage of structural sparsity assumption and highlight an implicit regularization phenomenon where the SGD algorithm with zeroth-order information adapts to the sparsity of the problem at hand by just varying the step-size. We next focus on avoiding saddle-points by utilizing Gaussian smoothing technique for estimating the gradient as an instantiation of first-order Stein’s identity. Based on this, we provide a novel linear-(in dimension) time estimator of the Hessian matrix using only noisy function evaluations, which is based on second-order Stein’s identity. We then provide a zeroth-order variant of cubic regularized Newton method for avoiding saddle-points and discuss its rate of convergence to local minima.

Finally, we will briefly discuss a new approach to solve inventory related problems, namely, an energy storage problem, under the presence of rolling forecasts and show how our proposed algorithms can be used to efficiently solve this kind of problem.



Speaker Bio

Saeed Ghadimi is an associate research scholar in the department of Operations Research and Financial Engineering at Princeton University. Saeed received his Ph.D. with major in Industrial and Systems Engineering from University of Florida. His research interests lie in the broad area of decision-making under uncertainty with an emphasize on nonconvex stochastic optimization motivated by data science applications.



The Operations Research (OR) seminar series brings together graduate students, faculty and researchers from the University of Toronto community to interact with prominent scholars in the field of OR. Seminars feature visiting scholars from around the world as well as professors and post-docs. Topics include all variants of OR theory and their applications. Questions? Contact Merve Bodur at bodur@mie.utoronto.ca