Abstract—We examine a simple biologically-motivated neural network, the version of the Chialvo-Bak “minibrain”, and propose an approach to decrease the negative effect of the active paths interferences in a process of learning new data. For this purpose we use randomly ordered neural network structure with recurrent signal propagation mode. We investigated the network's performance and learning capacity dependence on its nodes' interconnection level. Our simulation study shows that the proposed approach needs on average 40% less number of learning steps for learning the same set of patterns and has higher learning capacity compared to the existing method.
Index Terms—Hebbian learning, neural network, pattern recognition, reinforcement learning.
Anton Kulakov and Mark Zwoliński are with the University of Southampton, Southampton, UK (e-mail: ak06r@ ecs.soton.ac.uk; mz@ ecs.soton.ac.uk).
Cite: A. Kulakov and M. Zwoliński, "Reducing the Active Paths Interference in the Chialvo-Bak “Minibrain” Model," International Journal of Modeling and Optimization vol. 2, no. 6, pp. 734-737, 2012.
Copyright © 2008-2024. International Journal of Modeling and Optimization. All rights reserved.
E-mail: ijmo@iacsitp.com