The Price of Fair PCA: One Extra Dimension

Abstract

We investigate whether the standard dimensionality reduction technique of PCA inadvertently produces data representations with different fidelity for two different populations. We show on several real-world data sets, PCA has higher reconstruction error on population A than on B (for example, women versus men or lower- versus higher-educated individuals). This can happen even when the data set has a similar number of samples from A and B. This motivates our study of dimensionality reduction techniques which maintain similar fidelity for A and B. We define the notion of Fair PCA and give a polynomial-time algorithm for finding a low dimensional representation of the data which is nearly-optimal with respect to this measure. Finally, we show on real-world data sets that our algorithm can be used to efficiently generate a fair low dimensional representation of the data.

Cited by in corpus (17)

• FACT: A Diagnostic for Group Fairness Trade-offs
• Equalized odds postprocessing under imperfect group information
• Fair Principal Component Analysis and Filter Design
• On Tilted Losses in Machine Learning: Theory and Applications
• Fairness in Forecasting and Learning Linear Dynamical Systems
• Analysis of Trade-offs in Fair Principal Component Analysis Based on Multi-objective Optimization
• Active Sampling for Min-Max Fairness
• Fair Clustering with Multiple Colors
• Adversarial Reweighting for Speaker Verification Fairness
• Principal Fairness: Removing Bias via Projections
• Tilted Empirical Risk Minimization
• Regularity Normalization: Neuroscience-Inspired Unsupervised Attention across Neural Network Layers
• A Survey of Learning Criteria Going Beyond the Usual Risk
• A Distributed Fair Machine Learning Framework with Private Demographic Data Protection
• Fair Kernel Regression via Fair Feature Embedding in Kernel Space
• Towards Integrating Fairness Transparently in Industrial Applications
• Pairwise Fairness for Ordinal Regression