Mapping Approximation by the GMR Neural Network

-The Generalised Mapping Regressor (GMR) is an incremental self-organizing neural network with adaptive chains (linking) among neurons. These chains yield supplementary information to the network. GMR is capable to approximate every function or relation (general mapping) and, simultaneously, its inverse function, if it exists, or the inverse relation. It also outputs all the solutions (even infinite), their corresponding mapping branches and, if the case, the equilevel surfaces. The basic principle is the transformation of the function approximation problem into a pattern recognition problem under an unsupervised framework. Some examples conclude the paper. Key-Words: branch detection, function approximation, inverse problems, neural networks, pattern recognition, self organization