Summary of Reinformer: Max-return Sequence Modeling For Offline Rl, by Zifeng Zhuang et al.
Reinformer: Max-Return Sequence Modeling for Offline RL
by Zifeng Zhuang, Dengyun Peng, Jinxin Liu, Ziqi Zhang, Donglin Wang
First submitted to arxiv on: 14 May 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 A novel approach to offline reinforcement learning (RL) has been developed, which integrates the goal of maximizing returns into existing sequence models. The proposed Reinforced Transformer (Reinformer) combines RL objectives with sequence modeling techniques to predict optimal actions and improve trajectory stitching capability. Empirically, Reinformer demonstrates competitive performance with classical RL methods on the D4RL benchmark and outperforms state-of-the-art sequence models in trajectory stitching ability. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Offline reinforcement learning uses sequence models that condition on hindsight information. However, this approach overlooks the core objective of maximizing returns, which affects its ability to learn from sub-optimal data. A new concept called max-return sequence modeling aims to address this issue by integrating the goal of maximizing returns into existing sequence models. The proposed Reinforced Transformer (Reinformer) uses this approach and achieves competitive performance with classical RL methods. |
Keywords
» Artificial intelligence » Reinforcement learning » Transformer
