Summary of Evidential Graph Contrastive Alignment For Source-free Blending-target Domain Adaptation, by Juepeng Zheng et al.
Evidential Graph Contrastive Alignment for Source-Free Blending-Target Domain Adaptation
by Juepeng Zheng, Yibin Wen, Jinxiao Zhang, Runmin Dong, Haohuan Fu
First submitted to arxiv on: 14 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- 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 This paper addresses a more realistic Domain Adaptation (DA) setting, known as Source-Free Blending-Target Domain Adaptation (SF-BTDA), where there is no access to source domain data and multiple target domains without labels. The authors propose Evidential Contrastive Alignment (ECA) to decouple the blending target domain and alleviate noisy target pseudo labels generated from the source model. To improve pseudo label quality, they develop a calibrated evidential learning module that iteratively improves accuracy and certainty, generating high-quality pseudo labels. Additionally, they design a graph contrastive learning method using domain distance matrices and confidence-uncertainty criteria to minimize distribution gaps in blended targets. The authors conduct experiments on three standard DA datasets and demonstrate significant performance gains for ECA compared to other methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making computers learn from different types of data without needing a lot of information upfront. It’s like trying to teach a kid new words by showing them pictures, but the kid has never seen those words before. The authors came up with a new way to make this process work better, using something called Evidential Contrastive Alignment. They tested their method on three different types of data and found that it performed much better than previous methods. This is important because it could help computers learn from lots of different sources of information, like pictures or videos, and make them more useful for things like self-driving cars or medical diagnosis. |
Keywords
» Artificial intelligence » Alignment » Domain adaptation
