Summary of Spurious Correlations in Machine Learning: a Survey, by Wenqian Ye et al.
Spurious Correlations in Machine Learning: A Survey
by Wenqian Ye, Guangtao Zheng, Xu Cao, Yunsheng Ma, Aidong Zhang
First submitted to arxiv on: 20 Feb 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 Machine learning systems can be misled by irrelevant features that accidentally match label patterns. This issue, known as spurious correlations, undermines models’ ability to generalize and remain robust. To combat this problem, we review current approaches and categorize state-of-the-art methods for addressing spurious correlations. We also summarize existing datasets, benchmarks, and metrics to facilitate future research. The paper concludes by discussing recent progress and ongoing challenges in this area, providing valuable insights for researchers. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine learning models can get confused when they see things that don’t really matter. This is called a “spurious” connection between features and labels. It’s like if you trained a model to recognize dogs by looking at their ears, but it also learned to recognize cars based on the background trees. When the tree changes, the car looks different too! To avoid this problem, we’re reviewing how experts tackle spurious correlations today and what they’ve achieved. This will help researchers in the future. |
Keywords
* Artificial intelligence * Machine learning
