Chaotic Complex-Valued Associative Memory with Adaptive Scaling Factor Independent of Multi-values

In this paper, we propose a chaotic complex-valued associative memory with adaptive scaling factor independent from multi-values. This model is based on the conventional chaotic complex-valued associative memory with adaptive scaling factor that can realize dynamic associations of multi-valued patterns. In the conventional chaotic complex-valued associative memory with adaptive scaling factor, parameters of the chaotic complex-valued neuron model are automatically adjusted according to the internal state of neurons. In the conventional model, a multi-value pattern is expressed by assigning points at positions obtained by equally dividing a unit circle of the complex plane into S multiple values. It has been confirmed that almost same recall ability can be obtained as in the case of performing manual adjustment in the model for \(S=4,\ 6,\ 8\), but no study has been conducted for other cases. In addition, it is known that the optimum method of automatically adjusting parameters also differs depending on the value of S. In this study, we also conduct experiments at \( S = 10,\ 12,\ 14\) and 16, and propose a method to automatically adjust the parameters of the chaotic complex-valued neuron model independently from the value of S. Computer experiments were carried out and it was confirmed that automatic adjustment of parameters can be performed in the proposed model without depending on multi-values.