Training a generalized discrete Hopfield network with fuzzy learning rule

A Hopfield network, a type of recurrent neural network, may be used as a tool for classification by storing exemplars as memories. This paper describes a method of growing a hybrid network for use in classification of patterns which incorporates fuzzy membership in the training algorithm. The hybrid network consists of 3 networks in sequence with the middle network being a fully recurrent Hopfield style network which changes in size: starting out as a single neuron. The first network is a one layer feedforward network while the last network is simply a selector network which selects components from the terminal state of the recurrent network. Connection matrices are determined, using a modified Widrow-Hoff learning rule, such that the class exemplars are attracted to exemplars within the same class. An arbitrary element is then classified by the class of its attractor. Before training a membership value is calculated for each training pattern which is made use of during training.