Summary of Disentangled Unsupervised Skill Discovery For Efficient Hierarchical Reinforcement Learning, by Jiaheng Hu et al.
Disentangled Unsupervised Skill Discovery for Efficient Hierarchical Reinforcement Learning
by Jiaheng Hu, Zizhao Wang, Peter Stone, Roberto Martín-Martín
First submitted to arxiv on: 15 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Robotics (cs.RO)
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 A new method, Disentangled Unsupervised Skill Discovery (DUSDi), is proposed to learn reusable skills from unsupervised interaction with the environment. Existing methods often learn entangled skills that are challenging to chain together for downstream tasks. DUSDi decomposes skills into disentangled components that affect only one factor of the state space, allowing for efficient skill reuse and hierarchical Reinforcement Learning. A mutual-information-based objective is used to enforce disentanglement, and value factorization optimizes this objective efficiently. Evaluations in challenging environments show that DUSDi outperforms previous methods in applying learned skills to solve downstream tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary AI agents can learn new skills by playing games without being taught. But it’s hard to make these skills work together. A new way to discover skills, called Disentangled Unsupervised Skill Discovery (DUSDi), makes it easier to use these skills for future tasks. This method breaks down skills into smaller pieces that only affect one part of the game state, making it simpler to combine them and solve problems. |
Keywords
» Artificial intelligence » Reinforcement learning » Unsupervised
