Tackling Big-Data Challenges in Stochastic and Nonlinear Optimization
主 题: Tackling Big-Data Challenges in Stochastic and Nonlinear Optimization
报告人: Prof. Guanghui Lan, University of Florida
时 间: 2014-01-02 10:00-11:00
地 点: Room 09 at Quan Zhai, BICMR（主持人：文再文）
This talk focuses on the design and analysis of efficient algorithms to tackle the big-data challenges in optimization. The last several years have seen an unprecedented growth in the amount of available data. While nonlinear, especially convex programming models are important to extract useful knowledge from raw data, high problem dimensionality, large data volumes, and inherent uncertainty present significant challenges to the design of optimization algorithms. Aiming to attack some of these challenges, we introduce: i) a new class of stochastic approximation algorithms that can yield the optimal rate of convergence for solving different stochastic optimization problems. Some of these optimal rates were obtained for the first time in the literature; and ii) a new class of deterministic first-order algorithms that can converge optimally, require no structural information and do not rely on line search, based on level methods. To the best of our knowledge, no such uniformly optimal first-order methods have been studied before in the literature. Applications of these stochastic/deterministic algorithms will be studied. We will also briefly discuss some other related work and possible future research directions.