Memory effects of the multi-charge network model

We propose a spin glass model which has interactions characterized by P kinds of charges carried by N kinds of particles. When the densities of the particles are assumed to be Ising spin variables, this model becomes a spin glass model which has an opposite interactional sign to the Hopfield model with P memorized patterns. The number of local minimum states and remanent magnetization are studied to clarify the memory effects of the model. We find that, as P/N decreases, both quantities tend to the maximum possible values of the thermodynamic limit. (author)