Fuzzy Automata Induction using Construction Method

Abstract

Recurrent neural networks have recently been demonstrated to have the ability to learn simple grammars. In particular, networks using second-order units have been successfully at this task. However, it is often difficult to predict the optimal neural network size to induce an unknown automaton from examples. Instead of just adjusting the weights in a network of fixed topology, we adopt the dynamic networks (i.e. the topology and weights can be simultaneously changed during training) for this application. We apply the idea of maximizing correlation in the cascade-correlation algorithm to the second-order single-layer recurrent neural network to generate a new construction algorithm and use it to induce fuzzy finite state automata. The experiment indicates that such a dynamic network performs well.