Uncertainty-Aware PCA for Arbitrarily Distributed Data Modeled by Gaussian Mixture Models

TL;DR

This paper introduces an uncertainty-aware PCA method that effectively projects complex, non-normal multidimensional data modeled by Gaussian mixture models into low-dimensional space, preserving distribution details.

Contribution

It extends PCA to handle arbitrary distributions via GMMs and incorporates user-defined weights, improving distribution fidelity in low-dimensional projections.

Findings

Better preservation of distribution details in low-dimensional space

Flexible weighting of multiple distributions

Improved over traditional UAPCA in experiments

Abstract

Multidimensional data is often associated with uncertainties that are not well-described by normal distributions. In this work, we describe how such distributions can be projected to a low-dimensional space using uncertainty-aware principal component analysis (UAPCA). We propose to model multidimensional distributions using Gaussian mixture models (GMMs) and derive the projection from a general formulation that allows projecting arbitrary probability density functions. The low-dimensional projections of the densities exhibit more details about the distributions and represent them more faithfully compared to UAPCA mappings. Further, we support including user-defined weights between the different distributions, which allows for varying the importance of the multidimensional distributions. We evaluate our approach by comparing the distributions in low-dimensional space obtained by our method and UAPCA to those obtained by sample-based projections.