Analysis of Periodic Orbits in Cellular Binary Neural Networks
Hotaka Udagawa, Toshimichi Saito · 2021
This paper studies cellular binary neural networks characterized by local binary connection and the signum activation function. The networks are simple in structure, are easy to implement, and can generate various binary periodic orbits. In order to analyze the dynamics, we define two feature quantities that evaluate complexity and stability of the periodic orbits. Using the two feature quantities, we investigate complex dynamics of typical networks.