Summary of Piper: Primitive-informed Preference-based Hierarchical Reinforcement Learning Via Hindsight Relabeling, by Utsav Singh et al.
PIPER: Primitive-Informed Preference-based Hierarchical Reinforcement Learning via Hindsight Relabeling
by Utsav Singh, Wesley A. Suttle, Brian M. Sadler, Vinay P. Namboodiri, Amrit Singh Bedi
First submitted to arxiv on: 20 Apr 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 Piper, a novel approach to hierarchical reinforcement learning, leverages preference-based learning to learn a reward model and relabel higher-level replay buffers. This relabeling-based method mitigates non-stationarity in existing approaches, demonstrating impressive performance across challenging sparse-reward tasks. The primitive-in-the-loop approach generates feedback using environmental rewards, replacing human-in-the-loop methods. Primitive-informed regularization prevents degenerate solutions by conditioning high-level policies to generate feasible subgoals for low-level policies. Extensive experiments show Piper achieves over 50% success rates in robotic environments where other baselines fail. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Piper is a new way for robots and computers to learn from experience. It helps them make good decisions even when there’s no human around to tell them what to do. Piper uses something called preference-based learning, which lets it figure out what makes sense. This helps it avoid getting stuck in bad habits. The team tested Piper on lots of challenging tasks and found that it was really good at solving problems, especially when other approaches failed. |
Keywords
» Artificial intelligence » Regularization » Reinforcement learning
