A generalized addressing concept for correlative memory and neural networks

A generalized addressing concept is introduced. Associative recall from content-addressable memory is represented as decomposed on two stages—determination of generalized address of object in a system memory with subsequent generalized recall at this generalized address. It is shown that correlative memory systems and neural networks may be naturally treated in terms of this concept. Associative recall decomposition corresponds to decomposition of correlative system operational cycle and to introduction of additional layer in its network representation. Hierarchical correlative systems corresponding to multi-layer network representation are also considered.A generalized addressing concept is direct generalization of the usual computer-addressing concept and may serve as a link between conventional computers and neural networks.