Local (Hebbian) Learning

The non-learning (auxiliary) pRAMs define which actions of the central (learning) pRAMs are to be rewarded or penalised.

The auxiliary pRAMs receive both the input vector and the output of the learning pRAM so that all possible states can be defined.

Since the auxiliary pRAMs have a probabilistic output, complex reward/penalty actions can be taken.

More data

T G Clarkson, August 1995