Summary of Inverse Reinforcement Learning with Multiple Planning Horizons, by Jiayu Yao et al.
Inverse Reinforcement Learning with Multiple Planning Horizons
by Jiayu Yao, Weiwei Pan, Finale Doshi-Velez, Barbara E Engelhardt
First submitted to arxiv on: 26 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 The paper presents an inverse reinforcement learning (IRL) problem where experts plan under a shared reward function with different, unknown planning horizons. Existing IRL approaches struggle to identify a reward function due to the larger feasible solution set resulting from discount factors being unknown. To address this challenge, the authors develop algorithms that learn a global multi-agent reward function with agent-specific discount factors capable of reconstructing expert policies. The paper characterizes the feasible solution space for both algorithms and demonstrates the generalizability of the learned reward function across multiple domains. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper looks at a tricky problem in machine learning called inverse reinforcement learning, where experts are trying to achieve goals using different methods. The challenge is that we don’t know how these experts value time – some might prioritize short-term rewards, while others focus on long-term gains. To solve this problem, the authors create new algorithms that can learn a single reward function for all experts, taking into account their unique approaches. This can help us better understand and mimic the behavior of experts in different domains. |
Keywords
* Artificial intelligence * Machine learning * Reinforcement learning
