短期课程：Georgia Institute of Technology 蓝光辉教授到访我院并讲授短期课程

2019年6月27日至2019年6月28日，来自H. Milton School of Industrial and Systems Engineering Georgia Tech的蓝光辉教授在我校为我院及相关院系的师生们讲授了以“Selection Models, Treatment Effects, and the Econometric Evaluation of Policy Design”为主题的系列课程。

首先，蓝老师在明德商学楼0102教室为大家简要介绍了随机优化（Stochastic optimization）的最新境况，以及常用的几种随机优化模型。随机优化目前已成为现代机器学习的基础，目前广泛应用于训练机器学习模型的常用随机优化算法包括随机梯度下降（stochastic gradient descent (SGD)）、随机镜像下降（stochastic mirror-descent）、加速随机梯度下降（accelerated stochastic gradient descent (SGD with momentum)）、非凸随机梯度及其加速度器（nonconvex SGD and its acceleration）、方差减少技术（variance reduction techniques）、随机增量梯度法（randomized incremental gradient methods）和分布式随机算法（distributed stochastic algorithms）等。

接着，蓝老师又在明德商学楼0102教室为大家讲授了在应用这些算法时所面临的常见问题，以及算法自身的发展与挑战。例如，与普通的BGD算法相比，SGD算法是从样本中随机抽出一组，训练后按梯度更新一次，然后再抽取一组，再更新一次。在样本量大的情况下，可以在不用训练完所有的样本的情况下，获得一个损失值在可接受范围之内的模型。但是由于单个样本的训练可能会带来很多噪声，SGD并不是每次迭代都向着整体最优化方向，这些问题也亟待开发者们解决。最后，蓝老师也向同学们介绍了如何加强各种算法的适用性。

参与此次课程的各位老师同学也和蓝老师就课程的内容、问题以及在现实中的应用等方面展开了热烈的讨论，课程取得了圆满成功。

附：蓝光辉教授简介

Guanghui (George) Lan is an A. Russell Chandler III Early Career Professor and Associate Professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech.

Dr. Lan`s research interests lie in theory, algorithms, and applications of stochastic optimization and nonlinear programming. Most of his current research concerns the design of efficient algorithms with strong theoretical performance guarantees and superior practical performance for solving challenging optimization problems. Dr. Lan is actively pursuing the application of stochastic and nonlinear optimization models/algorithms in machine learning/intelligence.

Dr. Lan received his Ph.D. from Georgia Tech in 2009 and served as a faculty member in the Department of Industrial and Systems Engineering at the University of Florida from 2009-2015. His research has been supported by the National Science Foundation (NSF), the Office of Naval Research (ONR) and Army Research Office.

His academic honors include an NSF CAREER Award, first place in the INFORMS JFIG Paper Competition, finalist in the Mathematical Optimization Society Tucker Prize, second place in the INFORMS George Nicholson prize, and first place in the INFORMS Computing Society Student Paper competition.

Dr. Lan serves as an associate editor for Computational Optimization and Applications (2014-present), Mathematical Programming (2016-present) and SIAM Journal on Optimization (2016-present).

E-mail:

george.lan@isye.gatech.edu

Homepage：

https://www.isye.gatech.edu/users/george-lan