Fast and slow learning in a neuro-computational model of category acquisition

Abstract

We present a neuro-computational model that, based on brain principles, succeeds in performing a category learning task. In particular, the network includes a fast learner (the basal ganglia) that via reinforcement learns to execute the task, and a slow learner (the prefrontal cortex) that can acquire abstract representations from the accumulation of experiences and ultimately pushes the task level performance to higher levels.