Optimizing neuronal complexity using wavelet based multiresolution analysis for Type-I fuzzy neural networks

In neural network the connection strength of each neuron is updated through learning. Through repeated simulations of crisp neural network, we propose the idea that for each neuron in the network, we can obtain reduced model with more efficiency using wavelet based multiresolution analysis (MRA) to form wavelet based quasi fuzzy weight sets (WBQFWS). Such type of WBQFWS provides good initial solution for training in type-I fuzzy neural networks thus the search space for synoptic connections is reduced significantly, resulting in fast and confident learning of fuzzy neural networks. As real data is subjected to noise and uncertainty, therefore, WBQFWS may be helpful in the simplification of complex problems using low dimensional data sets. Key-Words: crisp neural networks, wavelets based multiresolution analysis (MRA), extraction of fuzzy rules, fuzzy neural network (FNN), wavelet based quasi fuzzy weight sets (WBQFWS), density estimation techniques.)