Self-organization and associative memory. (Q1210715)

This second edition [for the first edition (Berlin 1984) see Zbl 0528.68062)] is an extended revision of a work which treats the basic problems of memory (representation, storage and selective recall) starting from the adaptive network concept used as memory model; this concept may be realized as a physical system which processes continuous signals and whose structure and parameters are also continuously modifiable. After a detailed survey of various problems related to the memory (such as: the purpose and nature of biological memory, basic mechanisms, elementary operations, and abstract aspects of functionality) the presentation of the classical learning systems and a systematic approach of memory mechanisms as adaptive filters follow; further on the possibilities offered by the two fundamental properties of the adaptive network - self-organization and associativity - are minutely presented. After these basic topics: (i) the solving of the pattern recognition problem using adaptive networks is analysed, (ii) the problem of biological memory is considered in the context of the acquired knowledge about information processing principles of adaptive networks, (iii) the topic of neural computing is addressed, and (iv) a survey on optical associative memories is presented. These last two topics are the major contribution to the second edition; they are motivated by the recent strong interest in neural networks as a new paradigm for computer structures (iii) and the boom of new hardware technologies for distributed associative memories (iv). There are plenty of models with the associated mathematical formalisms. The rich bibliography is updated.