Decision-Making in Reinforcement Learning

  • 2019-06-01 02:36:42
  • Arsh Javed Rehman, Pradeep Tomar
  • 1


In this research work, probabilistic decision-making approaches are studied,e.g. Bayesian and Boltzmann strategies, along with various deterministicexploration strategies, e.g. greedy, epsilon-Greedy and random approaches. Inthis research work, a comparative study has been done between probabilistic anddeterministic decision-making approaches, the experiments are performed inOpenAI gym environment, solving Cart Pole problem. This research work discussesabout the Bayesian approach to decision-making in deep reinforcement learning,and about dropout, how it can reduce the computational cost. All theexploration approaches are compared. It also discusses about the importance ofexploration in deep reinforcement learning, and how improving explorationstrategies may help in science and technology. This research work shows howprobabilistic decision-making approaches are better in the long run as comparedto the deterministic approaches. When there is uncertainty, Bayesian dropoutapproach proved to be better than all other approaches in this research work.


Quick Read (beta)

This feature is not avaialbe for this paper.