Summary
International Symposium on Nonlinear Theory and its Applications
2017
Session Number:B1L-B-2
Session:
Number:B1L-B-2-3
A Study on Synapse Update of Inactive Cells in Cortical Learning Algorithm
Takeru Aoki, Keiki Takadama, Hiroyuki Sato,
pp.391-394
Publication Date:2017/12/4
Online ISSN:2188-5079
DOI:10.34385/proc.29.B1L-B-2-3
Summary:
Cortical learning algorithm (CLA) is a time-series data prediction algorithm based on the behavior of human neocortex. CLA has many cells connected by synapses, receives a time-series data and predicts the data coming next while updating synapse network. The conventional CLA only update synapses of active cells contributed to the prediction during the learning, and other synapses of inactive cells are neglected and not updated. To encourage the synapse network construction and improve the prediction accuracy of CLA, in this work we propose methods to update synapses of inactive cells and verify its effectiveness on test time-series data with/without noise.