Solving Survival Gridworld Problem Using Hybrid Policy Modified Q-Based Reinforcement
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This paper explores a model-free value-based approach for solving survival gridworld problem. Survivalgridworld problem opens up a challenge involving taking risks to gain better rewards. Classic value-basedapproach in model-free reinforcement learning assumes minimal risk decisions. The proposed method involvesa hybrid on-policy and off-policy updates to experience roll-outs using a modified Q-based update equationthat introduces a parametric linear rectifier and motivational discount. The significance of this approach is itallows model-free training of agents that take into account risk factors and motivated exploration to gainbetter path decisions. Experimentations suggest that the proposed method achieved better exploration and pathselection resulting to higher episode scores than classic off-policy and on-policy Q-based updates.
Abstract
This paper explores a model-free value-based approach for solving survival gridworld problem. Survivalgridworld problem opens up a challenge involving taking risks to gain better rewards. Classic value-basedapproach in model-free reinforcement learning assumes minimal risk decisions. The proposed method involvesa hybrid on-policy and off-policy updates to experience roll-outs using a modified Q-based update equationthat introduces a parametric linear rectifier and motivational discount. The significance of this approach is itallows model-free training of agents that take into account risk factors and motivated exploration to gainbetter path decisions. Experimentations suggest that the proposed method achieved better exploration and pathselection resulting to higher episode scores than classic off-policy and on-policy Q-based updates.
참고문헌 (6)
- Vincent François-Lavet, "An Introduction to Deep Reinforcement Learning", Foundations and Trends® in Machine Learning, 11(3-4), pp.219~354 (2018), Now Publishers[DOI]
- Hessel, Matteo, "Rainbow: Combining Improvements in Deep Reinforcement Learning", The 32nd AAAI Conference on Artificial Intelligence, pp.3215~3222 (2018)
- Mnih, Volodymyr, "Playing Atari with Deep Reinforcement Learning", NIPS Deep Learning Workshop 2013 (2013)
- Leike, Jan, "AI Safety Gridworlds", arXiv preprint arXiv：1711.09883v2, ~ (2017)
- Sutton, R. S., "Reinforcement Learning: An introduction" (2018), The MIT Press
- Mnih, Volodymyr, "Asynchronous Methods for Deep Reinforcement Learning", arXiv preprint arXiv: 1602.01783v2, ~ (2016)