Interacting dreaming neural networks

Abstract We study the interaction of agents, where each one consists an associative memory neural network trained with same patterns and possibly different reinforcement-unlearning dreaming periods. Using replica methods, we obtain rich equilibrium phase diagram coupled agents. It shows phases such as student–professor phase, only benefits from while other is unaffected; a mutualism both benefit; indifferent insufficient neither are benefited nor impaired; amensalism unchanged damaged. In addition to paramagnetic spin glass phases, there also call reinforced delusion agents concur without having finite overlaps patterns. For zero coupling constant, model becomes reinforcement removal model, which Hopfield model. single pattern, it Mattis version Ashkin–Teller