报告题目：Best Subset Selection in Regression Models

报告人：Sarat Babu Moka 博士（澳大利亚新南威尔士大学）

报告时间：2024年7月21日（周日）上午 10:00-11:30

报告地点：数理楼一楼235报告厅

报告摘要：We consider the problem of best subset selection in regression models, where the goal is to find for every model size k, a subset of k features that best fits the response. This is particularly challenging when the total available number of features is very large compared to the number of data samples. I will introduce COMBSS, a novel continuous optimization method that identifies a solution path, a small set of models of varying size, that consists of candidates for the best subset in the model. COMBSS turns out to be fast, making subset selection possible when the number of features is well in excess of thousands. I will present simulation results to show its performance in comparison to the existing popular methods including Lasso, Mixed Integer Optimization, and Forward Stepwise. Because of the overall outstanding performance COMBSS, framing the best subset selection challenge as a continuous optimization problem opens new research directions for feature extraction for a large variety of regression models.

报告人简介：Sarat Babu Moka，澳大利亚新南威尔士大学教师、澳大利亚麦考瑞大学荣誉研究员。他于2017年在印度塔塔基础科学研究所获博士学位，2017-2021年在澳大利亚昆士兰大学进行博士后研究，2021-2023年担任澳大利亚麦考瑞大学研究员。他的研究领域广泛，包括应用概率、计算统计、机器学习等。他为高维环境下高效模型选择的优化方法做出了贡献，并开发了用于空间点过程和随机图的快速无偏抽样和估计技术。此外，他的研究重点延伸到深度神经网络的高效修剪方法，在《Nature Communications》、《Bernoulli》、《Statistics and Computing》等期刊发表多篇高水平论文。除了研究，他还积极教授统计学和深度学习课程，同时他参与写作了一本关于深度学习的专著。