Performance analysis of a new updating rule for TD(λ) learning in feedforward networks for position evaluation in Go game

In this paper, a new updating rule for applying temporal difference (TD) learning to multilayer feedforward networks is derived. Networks are trained to evaluate Go board positions by TD(λ) learning with different values of λ. Performance of each network is estimated by letting it play against other networks. Results show that nonzero λ gives better learning for the network and statistically, larger λ gives better performance