A Group of Gain-Comparing Automata

The behavior of large groups of identical automata in random media is studied. The probability of a change of action for each automaton is determined by comparing the gains over a group of automata. For the case in which the group comprises all the automata, the author formulates the corresponding continuous automata and gives the systems of differential equations describing the behavior of infinitely large groups of such automata. It is shown that these groups tend to the optimum behavior in a number of types of random media. A machine experiment is described which shows, for a particular case, that the behavior of a finite number of finite comparing automata is quasioptimum.