Absorbing and ergodic discretized two-action learning automata
IEEE Transactions on Systems, Man and Cybernetics
Introduction to Reinforcement Learning
Introduction to Reinforcement Learning
The STAR automaton: expediency and optimality properties
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
A new class of ε-optimal learning automata
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
Learning the global maximum with parameterized learning automata
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
New class of P-model absorbing ε-optimal learning automata was presented in this paper. The proposed learning automaton, Discretized Generalized Stochastic Estimator (DGSE) learning automaton, not only possesses the characteristics of the Stochastic Estimator Reward-inaction (SERI ) learning automaton and the Discretized Generalized Pursuit Algorithm (DGPA) learning automaton, but also converges with a remarkable speed and accuracy. The asymptotic behavior of the DGSE algorithm is analyzed. Furthermore, we stick out the pitfalls in the proof of SERI algorithm, proved the proposed DGSE algorithm to be ε-optimal, and pointed out that this proof process could be applied to prove SERI algorithm. It's known that the SERI learning automaton is the fastest learning automaton up to now, whereas, the proposed DGSE learning automaton is much faster than the SERI learning automaton. A great number of experiments and simulations verified the propose DGSE learning algorithm is quite efficient when operating in P-model stationary environment.