Tuning of Fuzzy Neural Network Classifier

Fuzzy Sets, Neural Network, Fuzzy Pattern Recognition, Restricted Coulomb Energy, Fuzzy Neural Network.

The paper delas with a Fuzzy Restricted Coulomb Energy (FRCE) neural network. The network has the same architecture as a primitive RCE neural network: it consists of three layers: input layer X, hidden layer H (prototype layer) and output layer Y. There is a full set of connections between input and prototype layers, but only partial set of connections exits between hidden and output layers. Each neuron in the input layer represents a feature of an incoming pattern that the network must assign to some pattern class. Each neuron in the hidden layer contains information about an example of a learned class that occurs in the training data. In the output layer, each neuron corresponds to an individual class, i.e. it represents one category of patterns. The FRCE principle and it structure, learning an recalling algorithms and some experiments with FRCE are described in the paper.

The paper proposes a method of tuning fuzzy parameters in a fuzzy neural network. The network structure, modified off-line learning algorithm, recalling algorithm and using of the network as a pattern classifier of digitalized characters are described here.