On the Noise Model of Support Vector Machines Regression

摘要: Support Vector Machines Regression (SVMR) is a learning technique where the goodness of fit measured not by usual quadratic loss function (the mean square error), but different called Ɛ-Insensitive Loss Function (ILF), which similar to functions used in field robust statistics. The well justified under assumption Gaussian additive noise. However, noise model underlying choice ILF clear. In this paper use that and Gaussian, variance are random variables. probability distributions for will be stated explicitly. While work presented framework SVMR, it can extended justify nonquadratic any Maximum Likelihood or AP osteriori approach. It applies only ILF, much broader class functions.