Summary of Improved Regret Bound For Safe Reinforcement Learning Via Tighter Cost Pessimism and Reward Optimism, by Kihyun Yu et al.
Improved Regret Bound for Safe Reinforcement Learning via Tighter Cost Pessimism and Reward Optimism
by Kihyun Yu, Duksang Lee, William Overman, Dabeen Lee
First submitted to arxiv on: 14 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Optimization and Control (math.OC)
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 This paper proposes a model-based algorithm for safe reinforcement learning, tackling an episodic finite-horizon tabular constrained Markov decision process with unknown transition kernel and stochastic reward and cost functions. The novel cost and reward function estimators provide tighter cost pessimism and reward optimism, guaranteeing no constraint violation in every episode while achieving a regret upper bound of ((C – C_b){-1}H{2.5} S). This improves upon the best-known regret upper bound and nearly matches the regret lower bound when C- C_b=(H). The algorithm deduces cost and reward function estimators via a Bellman-type law of total variance, leading to tighter bounds on expected sum of variances of value function estimates. Numerical results demonstrate the computational effectiveness of the proposed framework. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps robots learn new skills safely by solving a complex problem called safe reinforcement learning. It proposes a new algorithm that makes sure the robot doesn’t break any rules while trying to find the best way to do something. The algorithm uses special equations to estimate how much cost and reward the robot will get, and it does this in a way that is really efficient and effective. |
Keywords
» Artificial intelligence » Reinforcement learning
