报告题目：Kolmogorov-Smirnov learning by neuron networks with a nonconvex surrogate loss
报 告 人：方方 (华东师范大学 教授)
报告时间：2023-12-13(星期三) 上午10:00开始
报告地点：腾讯会议，会议号：554-982-032
报告摘要：Kolmogorov-Smirnov (KS) statistic has been widely used in many areas to evaluate the performance of binary classification. However, almost no classification algorithm tries to optimize it directly at the training stage due to the computational and theoretical challenges brought by the special form of KS. In this paper, we propose a novel Kolmogorov-Smirnov neural Network (KSNet) using KS as the optimization objective. The difficulty of non-smoothness of the empirical KS is overcame by introducing a smooth nonconvex surrogate function. KSNet brings great potential to improve the KS in test data especially for imbalanced data and it shows inspiring robustness to data noise. Theoretically, we establish the non-asymptotic excess risk bound of KSNet with a ReLU activated feedforward neural network and show its Bayes-risk consistency. Further we alleviate the curse of dimensionality by assuming that the input data is supported on a compact low-dimensional manifold. Experiments on a variety of real datasets confirm the advantages of KSNet over a lot of existing methods.
专家简介：方方，华东师范大学统计学院教授，博士生导师。本科和博士先后毕业于北京大学数学系和美国威斯康星大学统计系。在2013年加入华东师范大学之前，曾在通用电气金融集团和上海浦东发展银行任职多年。主要研究方向为缺失数据、模型平均、碎片化数据分析、KS学习。在包括 AOS/JOE/Biometrika在内的国际一流统计期刊上发表论文30余篇。先后主持和参与国家和省部级项目12项。目前主持国家自然科学基金重点项目“大数据背景下不完全数据的统计分析方法、理论和应用”。授权专利6项。曾获上海市自然科学二等奖。担任全国工业统计学教学研究会常务理事、中国现场统计研究会机器学习分会常务理事、数字经济与区块链技术分会副理事长，IMS China委员会委员，SCI期刊 Journal of Nonparametric Statistics 副主编。在应用领域长期关注信用评分和民航QAR大数据分析。出版统计科普小说《统计王国奇遇记》。
