Discovering interpretable classification rules from neural processed data

In this paper we describe a neuro-fuzzy model to extract interpretable classification rules from examples. Such model is trained in a parameter subspace where a number of formal properties, which characterize understandable knowledge bases, are satisfied. To deal with the curse of dimensionality problem, which occurs when our model is used in high-dimensional classification tasks, an "A Priori Pruning" method is also proposed.