Object Generation with Neural Networks (When Spurious Memories are Useful)

Object generation constitutes a new class of problems which can be solved using neural networks. It is reverse to classification and makes a good use of the stable network modes generally considered as undesired (spurious memories). Single-class networks are considered as basic elements instead of multiclass ones, and attractors of the former are treated as potential objects of the corresponding class. Development of multiple attractors reflects the network generalization abilities and converts the single-class network into the active generator of templates. Object generating networks can also be used as moduli of recognition systems which are free from some disadvantages inherent in multi-class discriminant networks. Copyright 1996 Elsevier Science Ltd.