Accurate Bayesian Data Classification without Hyperparameter Cross-validation

TY - JOUR

T1 - Accurate Bayesian Data Classification without Hyperparameter Cross-validation

AU - Coolen, ACC

AU - Sheikh, Mansoor

PY - 2019/2/27

Y1 - 2019/2/27

N2 - We extend the standard Bayesian multivariate Gaussian generative data classier 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 classication 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.

AB - We extend the standard Bayesian multivariate Gaussian generative data classier 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 classication 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.

KW - Hyperparameters, evidence maximization, Bayesian classication, high-dimensional data

M3 - Article

SN - 0176-4268

JO - JOURNAL OF CLASSIFICATION

JF - JOURNAL OF CLASSIFICATION

ER -