Neurosolver: a step toward a neuromorphic general problem solver

In this paper, a neural network based on a model of a biological cortical column is presented. The cortical column has been found to play the fundamental role in information processing in the cerebral cortex. The model of the column presented in this work displays similar functionality. The stress is laid on the cooperative behavior of many artificial columns interconnected in a network. The network is capable of recording trajectories of time related events. Those recorded trajectories let the network use time dependencies to perform breadth-first searches. The device can solve stimulus-response type problems in a given domain and because of that is called Neurosolver. The Neurosolver can use a context update mechanism to perform dynamic searches. The Neurosolver represents the first step toward a neuromorphic general problem solver.<>