GAMES Webinar 2019 – 93期(深度学习可解释性专题课程) | 顾险峰（State University of New York at Stony Brook and Harvard University），王小倩（University of Pittsburgh）
【GAMES Webinar 2019-93期】
报告嘉宾：顾险峰，State University of New York at Stony Brook and Harvard University
报告题目： Geometric Views to GAN Models
Generative Adversarial Net (GAN) is a powerful machine learning model, and becomes extremely successful recently. The generator and the discriminator in a GAN model competes each other and reaches the Nash equilibrium. GAN can generate samples automatically, therefore reduce the requirements for large amount of training data. It can also model distributions from data samples. In spite of its popularity, GAN model lacks theoretic foundation. In this talk, we give a geometric interpretation to optimal mass transportation theory, explain the relation with the Monge-Ampere equation, and apply the theory for the GAN model.
Dr. David Xianfeng Gu got his B.S. in Computer Science from Tsinghua university in 1994, Ph.D. in Computer Science from Harvard University in 2002 supervised by the Fields medalist, Prof. Shing-Tung Yau. Currently, Dr. Gu is a tenured associate professor in Computer Science department, also affliated with applied mathematics department in the State University of New York at Stony Brook. Dr. Gu is also an affiliated professor in the Center of Mathematical Science and Applications of Harvard University.
Dr. Gu is one of the major funders of an emerging inter-disciplinary field: Computational Conformal Geometry, which combines modern geometry, topology theory with computer science. Dr. Gu and his collaborators laid down the theoretic foundations and systematically developed the computational algorithms, and applied conformal geometric method in many fields in engineering and medicine, such as Computer Graphics, Computer Vision, Visualization, Geometric Modeling, Networking, Artificial Intelligence, Medical Imaging and Computational Mechanics and so on.
Dr. Gu has published more than 300 papers in the top academic journals and conferences in the field of pure and applied mathematics, engineering and medical fields. Dr. Gu has won many academic awards, such as US NSF CAREER award in 2005, Morningsidea applied mathematics gold award in 2013 and so on.
报告嘉宾：王小倩，University of Pittsburgh
报告题目：Building Interpretable Neural Networks via Additive Model
In this talk, I will introduce new additive models to improve the interpretability of neural networks. I first design an interpretable regression model and derive the model convergence rate under mild conditions in the hypothesis space. Secondly, I introduce a new framework to improve the interpretability of general deep learning methods.
Xiaoqian Wang is a Ph.D. candidate at the University of Pittsburgh. Her research interests span across multidisciplinary areas of machine learning and bioinformatics, with work published in NeurIPS, ICML, KDD, IJCAI, AAAI, RECOMB, ECCB, etc. She will join Purdue University Computer Engineering as a tenure-track assistant professor in Fall 2019.
GAMES主页的“使用教程”中有 “如何观看GAMES Webinar直播？”及“如何加入GAMES微信群？”的信息；