POLYNOMIAL TIME RADIAL BASIS FUNCTION LEARNING SYSTEMS

A novel modification of the Radial Basis Function-type neural network. It can be used for robust and reliable brain-like learning in intelligent computer systems of the future. The devices are brain-like parallel computers that can be used for pattern classification (such as in medical diagnosis, image analysis, character recognition, robotics and financial and business analysis) and function approximation (useful in automatic control, robotics and financial and business analysis). The main charactristics of these devicesd that sets them apart from all other existing neural network learning systems are: 1. they can both design and train the neural network in polynomial time, which no other learning system can do; 2. they can learn much more quickly than others when used for on-line learning.