Deep Restricted Bayesian Network BESOM
We are developing a new machine learning system which we call the restricted Bayesian network BESOM, inspired by the computational models of the cerebral cortex. BESOM has the deep network structure like deep learning.
By restricting the model of the conditional probability tables, each step of inference and learning can be done in linear time with respect to the number of edges. Based on the working hypothesis that the cerebral cortex is a kind of Bayesian networks, we aim to reproduce various functions realized by the visual area and language area of the brain.