Unveiling and Mitigating Generalized Biases of DNNs through the Intrinsic Dimensions of Perceptual Manifolds

TL;DR

This paper introduces a geometric approach to analyze and mitigate biases in deep neural networks by examining the intrinsic dimensions of data manifolds, leading to improved fairness and performance.

Contribution

It proposes a novel intrinsic dimension regularization method that promotes fairer and more accurate DNNs by balancing the geometric complexity of class manifolds.

Findings

IDR reduces model bias in image recognition tasks.

IDR improves model performance while enhancing fairness.

Intrinsic dimensions of data manifolds correlate with DNN fairness.

Abstract

Building fair deep neural networks (DNNs) is a crucial step towards achieving trustworthy artificial intelligence. Delving into deeper factors that affect the fairness of DNNs is paramount and serves as the foundation for mitigating model biases. However, current methods are limited in accurately predicting DNN biases, relying solely on the number of training samples and lacking more precise measurement tools. Here, we establish a geometric perspective for analyzing the fairness of DNNs, comprehensively exploring how DNNs internally shape the intrinsic geometric characteristics of datasets-the intrinsic dimensions (IDs) of perceptual manifolds, and the impact of IDs on the fairness of DNNs. Based on multiple findings, we propose Intrinsic Dimension Regularization (IDR), which enhances the fairness and performance of models by promoting the learning of concise and ID-balanced class perceptual manifolds. In various image recognition benchmark tests, IDR significantly mitigates model bias while improving its performance.