Wednesday Sep 20, 12pm-1pm, 1116-E Klaus
ActiVis: Visual Exploration of Industry-Scale Deep Neural Network Models
Minsuk (Brian) Kahng
Advisor: Prof. Polo Chau
While deep learning models have achieved state-of-the-art accuracies for many prediction tasks, understanding these models remains a challenge. Despite the recent interest in developing visual tools to help users interpret deep learning models, the complexity of models and large-scale datasets used in industry pose unique design, visualization, and system challenges that are inadequately addressed by existing work. Through participatory design sessions with over 15 researchers and engineers at Facebook, we have designed, developed, and deployed ActiVis, a visual analytics system for interpreting industry-scale deep learning models and results. By tightly integrating multiple coordinated views, such as a neuron activation matrix view for pattern discovery and comparison, users can explore complex deep neural network models at both instance- and subset-level. ActiVis has been deployed on Facebook's machine learning platform.
Minsuk (Brian) Kahng is a fifth year Ph.D. student, advised by Prof. Polo Chau. His research focuses on building interactive visualization tools for helping people explore and interpret machine learning and data analytics systems. His work has been presented at premier conferences, such as VAST, InfoVis, VLDB, SDM, WWW, and SIGIR. He is a recipient of the NSF graduate research fellowship and received both my master's and bachelor's degrees at Seoul National University in South Korea. His website is at http://minsuk.com/