Summary of Domain Generalization with Small Data, by Kecheng Chen et al.
Domain Generalization with Small Data
by Kecheng Chen, Elena Gal, Hong Yan, Haoliang Li
First submitted to arxiv on: 9 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
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 work proposes a novel approach to domain generalization in situations where only limited samples are available. Instead of relying on deterministic models, it uses a probabilistic framework to learn a representation that is invariant across domains. The method extends the empirical maximum mean discrepancy (MMD) metric to measure the difference between mixture distributions, and introduces a new contrastive semantic alignment (CSA) loss function that encourages positive embeddings to be closer together while pushing negative ones apart. By combining these two components, the proposed method can align distributions at both global and local levels. Experimental results on three medical datasets demonstrate its effectiveness in situations with limited data. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about a new way to make machines understand information from different sources even when we don’t have much data. Instead of using models that are sure about what they know, this method uses ideas from probability theory to learn how to represent data in a way that doesn’t change no matter where the data comes from. The approach has two main parts: one that measures how similar the distributions of data are and another that makes sure the machine can tell different things apart. This helps the machine understand information better, especially when we don’t have much data. |
Keywords
* Artificial intelligence * Alignment * Domain generalization * Loss function * Probability
