Neural computing with structured information

Techniques to index and structurally relate the memories stored in a neural network are considered. Indexing may be done either by appending indexing keys to memories or by using portion of the memory itself to perform self-indexing. The first method is useful for easy recall and it can be applied to develop a database system. The second method defines a mechanism for sequential associative memory and for the sequential generation of a number of states. The usefulness of these techniques for optimization problems and knowledge representation is discussed. In particular, the use of neural networks in conjunction with the technique of octree representation for computer vision is described