Rough Set Theory based Neural Network Architecture

This paper proposes a novel method of combining rough concepts with neural computation. The proposed new rough neuron consists of, one lower bound neuron and another boundary neuron. The combination is designed in such a way that the boundary neuron deals only with the random and unpredictable part of the applied signal. Such architecture effectively prunes the search space for the respective constituent neurons based on the certain and uncertain behaviors. This division results in an improved rate of error convergence in the back propagation of the neural network along with an improved parameter approximation during the network learning process. Preliminary structures of the rough neural network along with some testing results have been presented. Further, performance comparisons with some of the prevalent designs have been done.