Genetic TD(/spl lambda/) learning algorithm for policy evaluation problems

In this paper, tabular TD(/spl lambda/) learning algorithm is combined with genetic algorithm (GA) to solve stochastic policy evaluation problems. Unlike conventional TD(/spl lambda/) algorithm which has fixed control parameters, the proposed genetic TD(/spl lambda/) algorithm makes use of GA to optimize the control parameters while evaluating stochastic policies. Simulated experiments on stochastic policy evaluation problems show that genetic TD(/spl lambda/) algorithms not only realize the auto-tuning of control parameters but also have improved performance.