Neural network design for J function approximation in dynamic programming

This paper will show that a new neural network design can solve an example of difficult function approximation problems which are crucial to the field of approximate dynamic programming(ADP). Although conventional neural networks have been proven to approximate smooth functions very well, the use of ADP for problems of intelligent control or planning requires the approximation of functions which are not so smooth. As an example, this paper studies the problem of approximating the J function of dynamic programming applied to the task of navigating mazes in general without the need to learn each individual maze. Conventional neural networks, like multi-layer perceptrons(MLPs), cannot learn this task. But a new type of neural networks, simultaneous recurrent networks(SRNs), can do so as demonstrated by successful initial tests. The paper also examines the ability of recurrent neural networks to approximate MLPs and vice versa.