Summary of Contrastive Representation For Data Filtering in Cross-domain Offline Reinforcement Learning, by Xiaoyu Wen et al.
Contrastive Representation for Data Filtering in Cross-Domain Offline Reinforcement Learning
by Xiaoyu Wen, Chenjia Bai, Kang Xu, Xudong Yu, Yang Zhang, Xuelong Li, Zhen Wang
First submitted to arxiv on: 10 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 proposed novel representation-based approach measures the domain gap by learning a contrastive objective from diverse transition dynamics in source and target domains. This approach recovers the mutual-information gap between transition functions without suffering from the unbounded issue of dynamics gaps. A data filtering algorithm is introduced to selectively share transitions from the source domain based on the contrastive score functions. Experimental results demonstrate superior performance, using only 10% of the target data to achieve 89.2% of the state-of-the-art methods’ performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Cross-domain offline reinforcement learning tries to use information from one place to help another. But when the places are very different, this can make things worse. Some people try to fix this by looking at how different the two places are, but that doesn’t always work. This new approach looks at what’s important in each place and finds a way to match them up better. It also picks out just the right information from one place to help the other. This helps it do better than others, even when it only has a little bit of information. |
Keywords
» Artificial intelligence » Reinforcement learning
