Summary of Improving Robustness to Multiple Spurious Correlations by Multi-objective Optimization, By Nayeong Kim et al.
Improving Robustness to Multiple Spurious Correlations by Multi-Objective Optimization
by Nayeong Kim, Juwon Kang, Sungsoo Ahn, Jungseul Ok, Suha Kwak
First submitted to arxiv on: 5 Sep 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 The proposed novel training method tackles the challenge of training an unbiased and accurate model from a dataset with multiple biases. The problem is complicated by the presence of multiple undesirable shortcuts during training, which can be exacerbated by attempts to mitigate one bias. The approach groups training data to induce different shortcuts and optimizes a linear combination of group-wise losses while adjusting their weights dynamically. This method, rooted in multi-objective optimization theory, encourages a minimax Pareto solution. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a way to train an unbiased model from a dataset with multiple biases. The challenge is that the biases can cause shortcuts during training, and trying to fix one bias might make another one worse. To solve this problem, the authors group the data so different groups create different shortcuts. Then, they adjust the weights of these groups dynamically to balance their effects. |
Keywords
» Artificial intelligence » Optimization
