Learning algorithms for stochastic automata acting in non-stationary random environments

A class of non-linear learning algorithms for stochastic automata acting in a non-stationary random environment is considered. The non-stationary random environment is assumed to be described by a regular Markov chain. Necessary and sufficient conditions for absolute expediency of the learning algorithms are derived. These algorithms are simulated and the results are compared. © 1974 Taylor & Francis Group, LLC.