A Temporal Difference GNG-Based Algorithm That Can Learn to Control in Reinforcement Learning Environments

This paper proposes a new reinforcement learning algorithm called TD-GNG that uses the Growing Neural Gas (GNG) network to deal with environments of large domains. The proposed algorithm is capable to reduce the dimensionality of the problem by aggregating similar states. In experimental comparison against tile-coding in mountain car and puddle world, the TD-GNG showed an increase in the generalization without loosing quality in the policy obtained.