Neural network with learning capability - has inputs processed by power series based transformation units

摘要

The neural network processing system has a base unit in the form of a node processor (60) that uses a number of stages (62) to effect a linear transformation of input values. The transformation is based upon a power series function. The units may be implemented either in hardware or in software. The transformation also involves a comparison of values with threshold values. When the elements are included in a network a two axis system is produced. The output product (64) is fedback to a direct learning module coupled to the input transformation modules. ADVANTAGE - Improves probability calculations using lineasr transformation elements.