Conference Paper 構造適応型Deep Belief Networkからの知識獲得に関する一考察

鎌田,真  ,  市村,匠

Deep Belief Network (DBN) has an deep architecture that can represent multiple features of input patterns hierarchically with pre-trained Restricted Boltzmann Machines (RBM). The model of DBN has an advantage of visualization or knowledge acquisition of the trained network because it is generative stochastic model. We have already proposed the adaptive learning method of DBN that can find an optimal number of hidden neurons and layers in the learning phase. In this paper, some considerations about the knowledge acquisition of the trained network by visualizing the activation of hidden neurons is discussed.
開催日:平成28年7月16日 会場:広島大学

Number of accesses :  

Other information