This talk explains recent our results on factorized asymptotic Bayesian (FAB) inference. FAB inference is a new Bayesian approximation inference for model selection, and has been applied to many latent variable models including mixture models, hidden Markov models, hierarchical mixture of experts, etc. The talk presents a basic principle of FAB inference by summarizing recent publications in KDD, NIPS, ICML, AISTATS, etc. Also, the talk shortly introduces several predictive analysis applications using FAB inference in NEC.
Ryohei Fujimaki (Ph.D.) received MS degree in aerospace engineering from University of Tokyo in 2006 and Ph.D. in 2010. He became the youngest research fellow ever in the history of NEC Labs. due to his business and R&D contributions, and is leading advanced analytics R&D teams in US, Japan and China. He has kept publishing academic papers in top conferences such as KDD, ICML, NIPS, AISTATS as well as he has developed many predictive analysis solutions with clients as a top industrial data scientist. He is a recipient of the Fuji-Sankei Business Eye (Advanced Technology) award 2015 in Japan.