Solutions to Parallel and Distributed Computing Problems: Lessons from Biological Sciences

4.6 STOCHASTIC LEARNING AUTOMATA

Learning systems have been shown to be an efficient tool to deal with a large number of engineering problems. They are information-processing systems whose architecture and behavior are inspired by the structure of biological systems (the organism is born with relatively little initial knowledge and learns actions that are appropriate through trials and errors). The vocabulary and the concepts associated with learning automata are borrowed from biology and psychology.

Learning systems are adaptive machines that interact with an environment (the problem to be solved) and that dynamically learn the optimal action that the environment offers. Thus, a learning automaton is connected in a feedback loop to the environment, where the input to one is the output to the other (see Fig. 4.7). At every iteration (or sampling period), the automaton chooses an action from a finite action set, on the basis of the probability distribution. The selected action causes the reaction of the environment, which in turn is the input signal for the automaton. The role of the environment is therefore to establish the relation between the actions of the automata and the signal received at its input. In an adaptive systems, the learning automaton recursively updates its probability distribution on the basis of the environment response.

Figure 4.7 Stochastic learning automaton. ...