Prior knowledge for fuzzy knowledge-based artificial neural networks from fuzzy set covering

Prior knowledge in a symbolic form can serve to initialize a knowledge-based neural network. We present a method for encoding fuzzy classification rules derived from a machine learning algorithm based on a fuzzy set covering framework. The inductive-bias of the encoding can be adjusted to allow further rule refinement and acquisition. We investigate the effect of these parameters, and show that the classification results correspond exactly to the prior knowledge.