Summary of Trained Models Tell Us How to Make Them Robust to Spurious Correlation Without Group Annotation, by Mahdi Ghaznavi et al.
Trained Models Tell Us How to Make Them Robust to Spurious Correlation without Group Annotation
by Mahdi Ghaznavi, Hesam Asadollahzadeh, Fahimeh Hosseini Noohdani, Soroush Vafaie Tabar, Hosein Hasani, Taha Akbari Alvanagh, Mohammad Hossein Rohban, Mahdieh Soleymani Baghshah
First submitted to arxiv on: 7 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 In this paper, researchers propose a novel method called Environment-based Validation and Loss-based Sampling (EVaLS) to enhance model robustness against spurious correlation. The method uses losses from an Empirical Risk Minimization-trained model to construct a balanced dataset of high-loss and low-loss samples, mitigating group imbalance in data. This approach significantly enhances robustness to group shifts when equipped with a simple post-training last layer retraining. EVaLS also eliminates the need for group annotation in validation data by using environment inference methods to create diverse environments with correlation shifts. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary To improve model performance and fairness, researchers developed a method called Environment-based Validation and Loss-based Sampling (EVaLS). This approach helps models work better on different groups of people, even if they don’t have the same attributes. It does this by using losses from an existing model to create a balanced dataset. This makes it easier for models to perform well across different groups. EVaLS is a fast and effective way to improve model performance without needing extra information about different groups. |
Keywords
» Artificial intelligence » Inference
