Summary of Fair4free: Generating High-fidelity Fair Synthetic Samples Using Data Free Distillation, by Md Fahim Sikder et al.
Fair4Free: Generating High-fidelity Fair Synthetic Samples using Data Free Distillation
by Md Fahim Sikder, Daniel de Leng, Fredrik Heintz
First submitted to arxiv on: 2 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 paper presents Fair4Free, a novel generative model that uses data-free distillation to generate synthetic fair data. The approach involves training a teacher model to create fair representations, then distilling the knowledge to a student model without requiring access to the original training dataset. The distilled model is used to generate fair synthetic samples, which are evaluated against state-of-the-art models in terms of fairness, utility, and synthetic quality. The results show that Fair4Free outperforms existing methods by 5% for fairness, 8% for utility, and 12% for synthetic quality on both tabular and image datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a new way to make fake data fair using a special kind of learning called distillation. The method starts with a teacher model that makes good fair representations, then uses this knowledge to create a smaller student model that can also make good fair representations without needing the original training data. The student model is used to generate new, fair synthetic samples that are better than current methods in three ways: fairness, usefulness, and how well they match real-world data. |
Keywords
» Artificial intelligence » Distillation » Generative model » Student model » Teacher model
