gmr: Gaussian Mixture Regression

Abstract

gmr is a Python library for Gaussian mixture regression (GMR). GMR is a regression approach that models probability distributions rather than functions. Hence, it is possible to model multimodal mappings. In GMR we first learn a joint probability distribution p(x, y) of input x and output y through expectation maximization (Dempster et al., 1977) and then compute the conditional distribution p(y|x) to make predictions. Thus, training is the same procedure as in a standard Gaussian mixture model (GMM).