Summary of Decision Making in Non-stationary Environments with Policy-augmented Search, by Ava Pettet et al.
Decision Making in Non-Stationary Environments with Policy-Augmented Search
by Ava Pettet, Yunuo Zhang, Baiting Luo, Kyle Wray, Hendrik Baier, Aron Laszka, Abhishek Dubey, Ayan Mukhopadhyay
First submitted to arxiv on: 6 Jan 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary In this paper, the authors introduce Policy-Augmented Monte Carlo Tree Search (PA-MCTS), a hybrid planning approach that combines action-value estimates from an outdated policy with online search using an up-to-date model of the environment. This approach is designed to tackle sequential decision-making under uncertainty in non-stationary environments, where the environment changes over time. The authors prove theoretical results showing conditions under which PA-MCTS selects the one-step optimal action and bound the error accrued while following PA-MCTS as a policy. They also compare and contrast their approach with AlphaZero and Deep Q Learning on several OpenAI Gym environments, showing that PA-MCTS outperforms these baselines in non-stationary settings with limited time constraints. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about helping computers make better decisions when the world around them changes. Right now, there are two main ways to do this: reinforcement learning and online search. Reinforcement learning is like training a computer to play chess by having it play many games. Online search is like using a map to find the best path. But both of these methods have problems when the environment changes. In this paper, the authors create a new way called Policy-Augmented Monte Carlo Tree Search (PA-MCTS) that combines the strengths of both methods. They show that PA-MCTS can make better decisions than other approaches in situations where the world is changing. |
Keywords
* Artificial intelligence * Reinforcement learning
