Fair4Free: Generating High-fidelity Fair Synthetic Samples using Data Free Distillation

TL;DR

Fair4Free introduces a data-free distillation approach to generate high-fidelity, fair synthetic data, improving fairness, utility, and quality over existing methods without needing access to original datasets.

Contribution

The paper proposes a novel data-free distillation method for generating fair synthetic data, applicable when original data is private or inaccessible.

Findings

Outperforms state-of-the-art models in fairness, utility, and synthetic quality.

Achieves 5% increase in fairness, 8% in utility, and 12% in synthetic quality.

Effective on both tabular and image datasets.

Abstract

This work presents Fair4Free, a novel generative model to generate synthetic fair data using data-free distillation in the latent space. Fair4Free can work on the situation when the data is private or inaccessible. In our approach, we first train a teacher model to create fair representation and then distil the knowledge to a student model (using a smaller architecture). The process of distilling the student model is data-free, i.e. the student model does not have access to the training dataset while distilling. After the distillation, we use the distilled model to generate fair synthetic samples. Our extensive experiments show that our synthetic samples outperform state-of-the-art models in all three criteria (fairness, utility and synthetic quality) with a performance increase of 5% for fairness, 8% for utility and 12% in synthetic quality for both tabular and image datasets.