A1036
Title: Sparse Bayesian CNN
Authors: Yongdai Kim - Seoul National University (Korea, South) [presenting]
Insung Kong - Seoul National University (Korea, South)
jinwon park - Seoul National University (Korea, South)
Abstract: A sparse Bayesian CNN model is proposed where a sparse prior is out on the filters in each layer. We develop an efficient MCMC algorithm and investigate how well the proposed Bayesian CNN selects filters.