Minimum k-cluster algorithm for hierachical reinforcement learning
暂无分享,去创建一个
Discovering useful subgoals while learning is important for hierarchical reinforcement learning.In this paper,a minimum k-cluster algorithm is proposed,which can extract subgoals from the set of trajectories collected online by clustering them.The result of experiment shows that the algorithm can find all subgoals quite efficiently and has better subgoal discovery performance than the diverse density and FD algorithms.