报告一：Ranking Preserving Nonnegative Matrix Factorization
报告内容：Nonnegative matrix factorization (NMF), a well-known technique to find parts-based representations of nonnegative data, has been widely studied. In reality, ordinal relations often exist among data,such as dataiis more related tojthan toq. Such relative order is naturally available, and more importantly, it truly reflects the latent data structure. Preserving the ordinal relations enables us to find structured representations of data that are faithful to the relative order, so that the learned representations become more discriminative. However, this cannot be achieved by current NMFs. In thispresentation,Dr. Wangmake the attempt towards incorporating the ordinal relations and propose a novel ranking preserving nonnegative matrix factorization (RPNMF) approach, which enforces the learned representations to be ranked according to the relations.Shederive iterative updating rules to solve RPNMF’s objective function with convergence guaranteed. Experimental results with several datasets for clustering and classification have demonstrated that RPNMF achieves greater performance against the state-of-the-arts, not only in terms of accuracy, but also interpretation of orderly data structure.
报告二：Brief talk about VRandAR
报告内容：In recent years，VR technology develop rapidly.The virtual realityis widely used in our daily life.Dr. Mawillintroduce how VR technology can change people's daily lives, especially in education, medical care, tourism, and others insights that have changed which could influenced us. And how we develop our own VR hardware software products. Similarly, we use VR an technologies together to develop new technology products that change traditional industries.
报告三：Hyperbolic Space for Hierarchical Data Embedding
报告内容：Embedding refers to a map from a object set to a space, with additional information preserved. Here, additional information includes an edge set in graph embedding setting and a distance matrix in metric multi-dimensional scaling. Although existing approaches have preserved such additional information in Euclidean space, whether Euclidean space is compatible with true data structure is largely ignored. which is essential to effective embedding. Since real data often exhibit hierarchical structure, it is hard for Euclidean space approaches to achieve effective embeddings in low dimensionality, which incurs high computational complexity or overfitting. Recent work has solved this problem by using hyperbolic space. In presentation,Mr. Suzukibriefly explain some basic properties of hyperbolic space, and how hyperbolic space works on embedding of hierarchical data.
报告人简介：王晶，现任日本东京大学博士后，研究领域包括机器学习，数据挖掘，专攻降维，聚类，多视角学习等。2018年1月于英国伯恩茅斯取得博士学位。在此之前，于香港城市大学取得多媒体资讯科技硕士学位。博期间以访问学生身份访问法国蒙彼利埃第二大学，美国纽约大学 和 以EU Marie Curie访问学者身份访问了澳大利亚查尔斯特大学。 博士期间凭借其研究成果获得全英“2017 ABTA Doctoral research award”工程自然科学类第二名。 目前共发表论文近20篇，其中包括CCF-A类会议AAAI 2019, IJCAI 2019，2018，2017，KDD 2019以及JCR-1区期刊IEEE transactions on Cybernetics, IEEE transactions on Image Processing等。并担任国际期刊及会议审稿人，包括AAAI 2020, 2019, TKDE, TIP, IEEE access等。
Atsushi Suzuki (铃木 惇), who is a third-year JSPS funded phd student in the graduate school of information science and technology of the University of Tokyo. Previous to that, he obtained both master and bachelor degrees from the same university. His research mainly focuses on deep learning, tensor factorization, information theory, statistics and so on.He has published several papers in top conferences, including ICDM, ISIT, AAAI, IJCAI, etc.