Events for 08-2019

Instructors/Speakers Prof. Jiebo LUO University of Rochester, the USA Abstract With the huge successes of deep learning in computer vision, many computer vision problems are seemingly being solved. Where do we go from here? We will discuss a few directions where computer vision can be either further pushed to deal with data scarcity and data noise, or synergistically integrated with other disciplines such as NLP and data mining, to continue to advance the frontiers of artificial intelligence. Biography Professor Jiebo Luo joined the University of Rochester in 2011 after a prolific career of fifteen years at Kodak Research Laboratories. He has been involved in numerous technical conferences, including serving as the program co-chair of ACM Multimedia 2010, IEEE CVPR 2012, ...

Instructors/Speakers Prof. Kaizhong ZHANG University of Western Ontario, London, Ontario, Canada Abstract Similarity and distance metrics are widely used in many research areas and applications. In some applications, similarity or distance metrics normalized with the "size" of the objects being measured are required. In this talk, we will first present a formal definition of similarity metric and then show general solutions to normalize a given similarity or distance metric. Examples and applications of the general solutions will also be presented. Biography K. Zhang received the M.S. degree in mathematics from Peking University, Beijing, China, in 1981, and the M.S. and Ph.D. degrees in computer science from the Courant Institute of Mathematical Sciences, New York University, New York, USA, in 1986 ...

Instructors/Speakers Prof. Min YANG Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences China Abstract Min Yang is currently an assistant professor at Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences. She received her Ph.D. degree from the department of computer science, the University of Hong Kong in 2017. Her current research interests include natural language processing, data mining, recommendation systems. Dr. Yang has more than 70 international, peer-reviewed publications on top-tier conferences or journals, such as ACL, SIGIR, WWW, KDD, AAAI, IJCAI, TKDE, TMM, etc. Biography a  

Instructors/Speakers Prof. Junchi YAN Shanghai Jiao Tong University China Abstract In this talk, I will first give a brief introduction on graph matching, which is a combinatorial problem in nature. Then we will show two deep network based pipelines for addressing the graph matching problem via deep learning. The models involve learning of the association based graph node embedding, cross-graph affinity learning, and a Sinkhorn layer for solving the linear assignment task, etc. We will also discuss some works on joint matching and link prediction among two or multiple graphs. In the end, some discussion will be given on the future work and outlook for connecting graph matching with machine learning. Biography Dr. Junchi Yan is currently an Independent Research ...

Instructors/Speakers Prof. Jingyi YU ShanghaiTech University China Abstract There have been tremendous advances on applying deep learning techniques for 2d image understanding. In contrast, very little work has focused on employing deep learning for modeling datasets beyond 2D such as 3D geometry and 4D light fields. In this talk, I present several latest works from our group on in this exciting new arena, with a focus on their applications to virtual and augmented reality and computational photography. I first present a novel deep surface light field (DSLF) technique. A surface light field represents the radiance of rays originating from any points on the surface in any directions. Traditional approaches require ultra-dense sampling to ensure the rendering quality. Our DSLF works ...

Instructors/Speakers Prof. Junliang XING Institute of Automation, Chinese Academy of Sciences China Abstract Face is perhaps the most important visual object in computer vision, with extensive studies in the past decades. In the deep learning era, the performances of computer vision problems related to faces have been significantly boosted, many of which have already met the requirements in real-world applications. In the talk, I will first make some basic introductions on the face vision problems, including face detection, face alignment, face tracking, face attribute analyses and face recognition. Then I will introduce some of our previous works related to this topic. At last, I will point out some future trends in this direction. The main objective of this talk is ...

Instructors/Speakers Prof. Jiwei ZHANG School of Mathematics and Statistics Wuhan University China Abstract Nonuniform time-stepping methods are promising for Caputo reaction sub-diffusion problems because they would be simple and effectiveness in resolving the initial singularity and other nonlinear behaviors occurred away from the initial time. Compared with traditional local methods for the first-order derivative, the numerical analysis for nonlocal time-stepping schemes on non-uniform time meshes are challenging due to the convolution integral (nonlocal) form of fractional derivative. We develop a general framework for the stability and convergence analysis with three tools: a family of complementary discrete convolution kernels, a discrete fractional Gronwall inequality (DFGI) and a global (convolutional) consistency analysis, which is not limited to a specific time mesh by ...

Instructors/Speakers Prof. Zhe XUE Beijing University of Posts and Telecommunications China Abstract Real-world data can be described from multiple views. For instance, an image can be described by color histogram, SIFT, HOG and other features. The content of a Web page can be described by texts, images and videos. Describing an object from multiple perspectives constitutes multi-view data. Multi-view learning is to use the complementary nature of different views to improve the learning performance than using a single view. In this talk, I will first briefly introduce some basic concepts and issues in multi-view learning. Then I will introduce some of our recent multi-view learning works for image analysis including multi-view dimensionality reduction, multi-view clustering and incomplete multi-view classification. This ...

Instructors/Speakers Prof. Fei Fang Institute for Software Research School of Computer Science Carnegie Mellon University USA Abstract There is a rising interest in developing artificial intelligence-based tools to help address societal challenges. Motivated by these challenges, we have proposed game theory and machine learning/reinforcement learning-based models and algorithms for problems with strategic interactions among agents. In this talk, I will introduce our models and algorithms that have led to two successfully deploy applications: one used by US Coast Guard for protecting the Staten Island Ferry in New York City since April 2013, the other used in multiple conservation areas around the world for anti-poaching effort. In addition, I will highlight our most recent advances in integrating deep learning with game ...

Instructors/Speakers Dr. Shuai-Xia XU Lecturer Sun Yat-sen University China Abstract We study the gap probabilities in the critical unitary invariant random matrix ensembles, where the Painleve II and Painleve XXXIV kernels arise. By studying the Fredholm determinants of the Painleve II and Painlev XXXIV kernels, we obtain integral expression of the gap probabilities by using solutions to the coupled Painleve II system. Moreover, the large gap asymptotics are derived with the constant terms given explicitly in terms of the Riemann zeta-function. This talk is based on a joint work with Dan Dai from City University of HongKong. Biography Prof. Xu Shuaixia got his PhD degree from Sun Yat-sen University in 2011. From 2011 to 2013, he worked as a postdoctoral ...