Evaluating Fairness and Mitigating Bias in Machine Learning: A Novel Technique using Tensor Data and Bayesian Regression

TL;DR

This paper presents a novel fairness evaluation and bias mitigation technique for machine learning models handling tensor data like skin color, using Bayesian regression to improve equity in image classification without relying on annotations.

Contribution

It introduces a new approach to assess and mitigate bias in ML models dealing with tensor data, specifically skin color, by converting data into probability distributions and applying Bayesian regression.

Findings

Effective fairness evaluation without annotations

Improved bias mitigation in skin tone classification

Captures nuanced differences across skin color groups

Abstract

Fairness is a critical component of Trustworthy AI. In this paper, we focus on Machine Learning (ML) and the performance of model predictions when dealing with skin color. Unlike other sensitive attributes, the nature of skin color differs significantly. In computer vision, skin color is represented as tensor data rather than categorical values or single numerical points. However, much of the research on fairness across sensitive groups has focused on categorical features such as gender and race. This paper introduces a new technique for evaluating fairness in ML for image classification tasks, specifically without the use of annotation. To address the limitations of prior work, we handle tensor data, like skin color, without classifying it rigidly. Instead, we convert it into probability distributions and apply statistical distance measures. This novel approach allows us to capture fine-grained nuances in fairness both within and across what would traditionally be considered distinct groups. Additionally, we propose an innovative training method to mitigate the latent biases present in conventional skin tone categorization. This method leverages color distance estimates calculated through Bayesian regression with polynomial functions, ensuring a more nuanced and equitable treatment of skin color in ML models.