PHOENIX, AZ: Dr. Guanghui (George) Lan, Assistant Professor of Industrial and Systems Engineering, and his Ph.D. student Saeed Ghadimi have been named finalists for the 2012 Institute for Operations Research and the Management Sciences (INFORMS) JFIG (Junior Faculty Interest Group) Paper Competition. The nomination is for their paper, “Stochastic First- and Zeroth-order Methods for Nonconvex Stochastic Programming,” which was completed this summer. Of the 50 entries, six were named finalists, with the winner to be named a the INFORMS annual conference in Phoenix this October.
This paper introduces novel randomized stochastic gradient methods for solving an important class of nonlinear, possibly nonconvex stochastic programming problems. New complexity bounds and strong large-deviation properties are presented for these algorithms, along with promising numerical results. Specialized stochastic zeroth-order algorithms are also developed for solving a class of simulation-based optimization problems. This work seems to be the first effort in developing nonconvex stochastic gradient methods in the literature.
The JFIG Forum of INFORMS was created in 2001 to promote the career development of tenure-track faculty in operations research. According to the INFORMS website, the paper competition is meant to encourage and promote research among junior faculty in the field.