An Architecture for Pattern Recognition and Decision-Making

We present a simple brain architecture that allows agents to recognise patterns and make decisions based on those patterns. It takes into account not only the type of situation the agent thinks it is facing, but also how confident the agent is in its assessment, and possible alternatives. An agent using this brain was applied to two classification tasks: handwritten numeral recognition and spoken numeral recognition. In both cases, its accuracy was comparable to more traditional classifiers. This suggests that the new architecture could be useful as a general-purpose brain, for agents in a variety of domains.