Analysis and synthesis of continuous‐time hysteretic neural networks
Kenya Jin’no, Toshimichi Saito · Electronics and Communications in Japan (Part III Fundamental Electronic Science) · 1993
Abstract In this article the authors propose a continuous‐time neural network which includes bipolar hysteretic elements. They then consider the problem of synthesis in order to satisfy the following conditions: I) the system has no oscillating solutions; II) any desired patterns are stored as stable output vectors; and III) the total number of false stable output vectors is made as small as possible. The authors then prove a sufficient condition for I) and the necessary and sufficient conditions for II). Based on these results, they propose a synthesis procedure which is also effective for III). The domain of attraction of stable outputs may be easily calculated using a fast algorithm using exact solutions. The efficiency of the procedure is confirmed through experimental results.