Accurate Bayesian Data Classification without Hyperparameter Cross-validation

TL;DR

This paper introduces an improved Bayesian data classifier that optimizes hyperparameters analytically, achieving competitive accuracy without the computational cost of cross-validation, especially effective in high-dimensional data scenarios.

Contribution

It generalizes the Bayesian Gaussian classifier with an analytically derived hyperparameter optimization, eliminating the need for cross-validation.

Findings

Achieves competitive classification accuracy with reduced computational effort.

Effectively handles high-dimensional data regimes.

Eliminates the need for hyperparameter cross-validation.

Abstract

We extend the standard Bayesian multivariate Gaussian generative data classifier by considering a generalization of the conjugate, normal-Wishart prior distribution and by deriving the hyperparameters analytically via evidence maximization. The behaviour of the optimal hyperparameters is explored in the high-dimensional data regime. The classification accuracy of the resulting generalized model is competitive with state-of-the art Bayesian discriminant analysis methods, but without the usual computational burden of cross-validation.