Summary
International Symposium on Nonlinear Theory and Its Applications
2016
Session Number:C1L-F
Session:
Number:C1L-F-4
Efficient Implementation of Boltzmann Machine Using Asynchronous Network of Cellular Automaton-Based Neurons
Takashi Matsubara, Kuniaki Uehara,
pp.-
Publication Date:2016/11/27
Online ISSN:2188-5079
Summary:
Artificial neural networks with stochastic state transitions, such as Boltzmann machines, have excelled other machine learning approaches in various benchmark tasks. They however require implementation of nonlinear continuous functions and generation of numerous pseudo random numbers, resulting in increase in computational resources. This study proposes a novel implementation method of Boltzmann machine using asynchronous network of cellular automaton-based neurons. The proposed approach requires much less computational resources than traditional implementation approaches since it does not require both the nonlinear continuous functions.