Integration of Artificial Neural Networks and Dynamic Concepts to an Adaptive and Self-Organizing Agent

A brain model based alternative to reinforcement learning is presented. It integrates artificial neural networks (ANN) and knowledge based (KB) systems into one unit or agent for goal oriented problem solving. The agent may possess inherited and learnt ANN and KB subsystems. The agent has and develops ANN cues to the environment for dimensionality reduction in order to ease the problem of combinatorial explosion. Here, a dynamic concept model is forwarded that builds cue-models of the phenomena in the world, designs dynamic action sets (concepts) and make them compete in a neural stage to come to a decision. The agent works under closed-loop control. Here we examine a simple robotic-like object in a two dimensional non-probabilistic world.