综合新闻
南京航天航空大学陈松灿教授受邀来研究院做学术报告
2021年12月4日,南京航天航空大学陈松灿老师,受邀来我校作报告。本次报告以腾讯会议的形式举行,由乔立山教授主持,数学科学学院部分教师、研究生聆听了报告。
陈松灿教授在第一个报告中,以模型选择是机器学习的核心。展开介绍了交叉验证(Cross-Validation-CV)作为种广泛采纳并在经验上行之有效的选择策略,常常面临需多重数据划分验证所导致的效率低下,以及在少标记样本学习场景下难奏效等问题,在本报告种给出一种无需留出验证集的模型选择新策路:Leave-Zero-Out,其不仅高效和有效,而且具有用适合监督、半监督到无监督模型选择的普适性。第二个报告中,陈松灿教授以Multi-label learning (MLL) is an important learning paradigm in machine learning where completing those missing labels is a key One of Popular MLLs works under the low-rankness assumption for Iabel completion, however, this is often violated in practice. In this talk, we first illustrate the high-rankness in single/multiview MLLS and more challenges in the latter setting, then provide the concise yet effective methods to meet them.会后师生与陈松灿教授对会议的内容进行了更加深入的探讨。