ABSTRACT
Biological systems routinely solve problems involving pattern recognition and feature extraction. A basic characteristic of biological systems is that they tend to evolve, through modes of information processing, toward "suitable" modes most appropriate to their environment. The dynamics of this process has been captured and indirectly exploited in simulated annealing, as used in attacking a variety of optimization problems. Whereas evolutionary learning is an essential attribute of biological intelligence, the ability to learn is by no means sufficient to account for pattern recognition, feature extraction, and, in the case of "sentient" intelligence, self-reference. By a history-dependent stochastic automaton (HDSA), mean a tuple (I,S,O,R,P), where: I is a set (of external inputs), including a null state;S is a (possibly infinite) set of internal states, including an initial state;O is a set (of outputs);R is a function from the set of states to the set of outputs; andP is a function which assigns to each history..
