This paper has presented a new representation for learning complex programs. This new representation, Neural Programming has been developed with the goal of incorporating positive aspects of both artificial neural networks and genetic programming. Neural Programming is a connectionist programming language which has been designed to make internal reinforcement, hither-to unaccomplished in genetic programming, possible. A simple way of accomplishing this sort of internal reinforcement was detailed and some alternatives and extensions were briefly mentioned. We illustrated the technique with a focused experiment that showed that internal reinforcement improves the speed and accuracy of Neural Programming learning.
Neural Programming and associated internal reinforcement policies are our on-going research. The goal of this paper has been to communicate the exciting possibility that, through the exploration of new program representations, we may be able to find ways to capture the explanation and update power of backpropagation with the flexibility and generality of genetic programming.