Summary of Stitching Sub-trajectories with Conditional Diffusion Model For Goal-conditioned Offline Rl, by Sungyoon Kim et al.
Stitching Sub-Trajectories with Conditional Diffusion Model for Goal-Conditioned Offline RL
by Sungyoon Kim, Yunseon Choi, Daiki E. Matsunaga, Kee-Eung Kim
First submitted to arxiv on: 11 Feb 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- 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 proposes a novel approach to Offline Goal-Conditioned Reinforcement Learning (Offline GCRL), which focuses on learning diverse goal-oriented skills from pre-collected behavior datasets without reward feedback. The authors leverage the conditional diffusion model to generate long-horizon plans for RL, addressing limitations of previous methods. They introduce SSD (Sub-trajectory Stitching with Diffusion), a model-based offline GCRL method that estimates target values from goal-relabeled datasets and generates future plans conditioned on target goals and values. This approach achieves state-of-the-art performance in standard benchmark tasks and demonstrates the ability to stitch suboptimal trajectories into high-quality plans. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about using artificial intelligence to learn new skills without getting rewarded or punished for each small step. It’s like learning a new language by listening to native speakers, rather than practicing individual words. The researchers use a special type of model that can generate plans for the future based on what we want to achieve. They show that this approach works really well and can even fix mistakes in old data to create better plans. |
Keywords
» Artificial intelligence » Diffusion » Diffusion model » Reinforcement learning
