Talk by Mavashi Sugiyama, RIKEN/University of Tokyo

Title: Machine learning from weak supervision --- Towards accu rate classification with low labeling costs. Abstract: Recent advan ces in machine learning with big labeled data allow us to achieve human-l evel performance in various tasks such as speech recognition, image unde rstanding, and natural language translation. On the other hand, there a re still many application domains where human labor is involved in the da ta acquisition process and thus the use of massive labeled data is prohib ited. In this talk, I will introduce our recent advances in classificatio n techniques from weak supervision, including classification from two set s of unlabeled data, classification from positive and unlabeled data, an d a novel approach to semi-supervised classification. Bio: Masashi S ugiyama received the PhD degree in Computer Science from Tokyo Institute of Technology, Japan in 2001. He has been Professor at the University ofTokyo since 2014 and concurrently appointed as Director of RIKEN Center for Advanced Intelligence Project in 2016. His research interests in clude theory, algorithms, and applications of machine learning. He (co) -authored several books such as Density Ratio Estimation in Machine Learn ing (Cambridge University Press, 2012), Machine Learning in Non-Station ary Environments (MIT Press, 2012), Statistical Reinforcement Learning (Chapman and Hall, 2015), and Introduction to Statistical Machine Learn ing (Morgan Kaufmann, 2015). He served as a Program Co-chair and Ge neral Co-chair for the Neural Information Processing Systems conference in2015 and 2016, respectively, and he will be a Program Co-chair for AIST ATS2019. Masashi Sugiyama received the Japan Society for the Promotion of Science Award and the Japan Academy Medal in 2017.