Summary of Multi-output Distributional Fairness Via Post-processing, by Gang Li et al.
Multi-Output Distributional Fairness via Post-Processing
by Gang Li, Qihang Lin, Ayush Ghosh, Tianbao Yang
First submitted to arxiv on: 31 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (stat.ML)
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 introduce a new post-processing technique for enhancing fairness in machine learning models with multiple outputs. The goal is to improve distributional parity, a task-agnostic measure of fairness, which is challenging to achieve when working with multi-output models like those used in multi-task/multi-class classification and representation learning tasks. To address this limitation, the proposed method uses an optimal transport mapping to shift model outputs towards their empirical Wasserstein barycenter across different groups. The approach also employs an approximation technique to reduce computational complexity and a kernel regression method for extending out-of-sample data processing. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper develops a new way to make machine learning models more fair by changing how they work after training. Right now, there are many ways to make models fairer, but most of these methods only work for models that can produce one answer or output at a time. This new method is special because it can be used with models that can do multiple tasks and classify things into different categories. It uses a technique called optimal transport mapping to move the model’s outputs towards what’s normal for each group, making it more fair overall. |
Keywords
» Artificial intelligence » Classification » Machine learning » Multi task » Regression » Representation learning
