Estimating the ratio of two probability densities from finitely many observations of the densities, is a central problem in machine learning and statistics. A large class of methods constructs estimators from binary classifiers which distinguish observations from the two densities. However, the error of these constructions depends on the choice of the binary loss function, raising the question of which loss function to choose based on desired error properties. In this work, we start from prescribed error measures in a class of Bregman divergences and characterize all loss functions that lead to density ratio estimators with a small error. Our characterization provides a simple recipe for constructing loss functions with certain properties, such as loss functions that prioritize an accurate estimation of large values. This contrasts with classical loss functions, such as the logistic loss or boosting loss, which prioritize accurate estimation of small values. We provide numerical illustrations with kernel methods and test their performance in applications of parameter selection for deep domain adaptation.

Estimating the ratio of two probability densities from finitely many observations of the densities is a central problem in machine learning and statistics with applications in two-sample testing, divergence estimation, generative modeling, covariate shift adaptation, conditional density estimation, and novelty detection. In this work, we analyze a large class of density ratio estimation methods that minimize a regularized Bregman divergence between the true density ratio and a model in a reproducing kernel Hilbert space (RKHS). We derive new finite-sample error bounds, and we propose a Lepskii type parameter choice principle that minimizes the bounds without knowledge of the regularity of the density ratio. In the special case of quadratic loss, our method adaptively achieves a minimax optimal error rate. A numerical illustration is provided.

The ratio of two probability densities can be used for solving various machine learning tasks such as covariate shift adaptation (importance sampling), outlier detection (likelihood-ratio test), and feature selection (mutual information). Recently, several methods of directly estimating the density ratio have been developed, e.g., kernel mean matching, maximum likelihood density ratio estimation, and least-squares density ratio fitting. In this paper, we consider a kernelized variant of the least-squares method and investigate its theoretical properties from the viewpoint of the condition number using smoothed analysis techniques--the condition number of the Hessian matrix determines the convergence rate of optimization and the numerical stability. We show that the kernel least-squares method has a smaller condition number than a version of kernel mean matching and other M-estimators, implying that the kernel least-squares method has preferable numerical properties. We further give an alternative formulation of the kernel least-squares estimator which is shown to possess an even smaller condition number. We show that numerical studies meet our theoretical analysis.