[Seminar] Visual Analytics for Interpretable and Interactive Deep Neural Networks
호스트: 서진욱 교수(x7044)
Interactive visualization is an effective means by which humans can obtain insights on the inner-workings of deep neural networks and steer these models in a user-driven manner. This talk will present some of my recent work along this line of research, including (1) SANVis, a visual analytic system for understanding self-attention networks in natural language processing, (2) RetainVis, another visual analytics system for interpretable and interactive time-series prediction using electronic medical records, and (3) AILA, an interactive document labeling interface via attention visualization and supervision. Afterwards, I will discuss recent research trends and opportunities in the areas of information visualization, visual analytics, and human-in-the-loop artificial intelligence.
Jaegul Choo (https://sites.google.com/site/jaegulchoo/ ) is an associate professor in the Dept. of Computer Science and Engineering at Korea University. He has been a research scientist at Georgia Tech from 2011 to 2015, where he received M.S in 2009 and Ph.D in 2013. His research areas include visual analytics, computer vision, natural language processing, and data mining, and and his work has been published in premier venues such as IEEE VIS, EuroVIS, CHI, TVCG, CVPR, ECCV, EMNLP, AAAI, IJCAI, KDD, WWW, WSDM, ICDM, ICWSM. He earned the Best Student Paper Award at ICDM in 2016, the NAVER Young Faculty Award in 2015, the Outstanding Research Scientist Award at Georgia Tech in 2015, and the Best Poster Award at IEEE VAST (as part of IEEE VIS) in 2014.