||Hanwang Zhang 張含望
||Nanyang Assistant Professor ，Nanyang Technological University, Singapore
||2019 年 12 月 27 日（周五）上午 10 點
||張江校區軟件樓 105 IBM 會議室
Today’s deep models in the vision-language community are too brittle when we ask “why”: Why is the model designed like this? or, Why do we need to add visual attention here? or, Why is this model better than that one? Over the years, you must be obsessed with these why-s. Don’t worry, it seems that everyone is happy with them. Maybe it’s because we are indulging in the pleasure of model design, or we are just too lazy to think about them. So, till now, no one explains those why-s. In this talk, I will be the one! I will introduce Causality --- the glory of mind --- to dispel the mysteries. To show its power, I will offer four case studies: Image Captioning, Visual Dialog, VQA, and Scene Graph Generation. By using causality, we can improve all these tasks significantly without the need for designing any new model. After the talk, I hope you know “how”.
Dr. Hanwang Zhang is Nanyang Assistant Professor at Nanyang Technological University, Singapore. He was a research scientist at the Department of Computer Science, Columbia University, USA. He has received the B.Eng (Hons.) degree in computer science from Zhejiang University, Hangzhou, China, in 2009, and the Ph.D. degree in computer science from the National University of Singapore in 2014. His research interest includes machine reasoning and its applications in computer vision, natural language processing, and multimedia.