Optimization and order reduction of networks of high order neurons

The concept of one shot learning is reviewed and a new method is developed. Concerns which arise from uses of various neural networks to which this new technique of decomposition is applicable, as well as advantages we believe can result from the use of the decomposition method are discussed. A procedure by which one can reduce the order of the neurons to the minimum which will successfully solve a problem and determine their weights which will guarantee monotonic convergence to the desired local minima is also developed. In order to demonstrate the usefulness of these new techniques, several classes of problems and previously applied solution methods are discussed