Chaotic dynamics applied to information processing

Abstract

Information processing aims at category formation in the (human) cognitive system or the classification of the external (and a good deal of the system's internal) world in a hierarchy of abstract patterns. These patterns in turn may be used as algorithms for simulating observed physical processes. In this review the author examines how this abstraction process may be dynamically implemented using the concepts of 'chaotic strange attractors'. The paper is divided accordingly into three parts. (1) Introduction, where the author deliberates on the physical (hardware) substrate of information and its hierarchical relation to the symbolic (software) aspect; or, more concretely, how from a continuous non-linear dynamics to obtain a Markov chain. (2) Where the dynamics of information generation, compression and dissipation is given. (3) Applications where examples of information processing using the paradigm of chaotic attractors are offered from neurophysiology, cognitive psychology and perception. Finally the author discusses the issue of self-consistency of self-referential linguistic schemes essentially as an eigenvalue problem.