A neural network for analogical reasoning

Analogical reasoning is one of the fundamental mental processes. It is a multi-step procedure that involves retrieval and manipulation of stored information, and generation of new inferences. Presented here is a model neural network, that encodes and stores information, so that it can be easily accessed in analogical reasoning processes. Learning rules that the network uses for incorporating new inferences are formulated. The interactions between the network and an external controller, that controls the execution of the various steps of the analogical reasoning process, are described, and an example that illustrates how these principles operate in a typical analogical reasoning problem is given. Some of the learning rules that the network employs are different from those currently utilized in other connectionist models of the brain.<
>