posted on 2017-12-29, 18:21authored byFlorian Wenzel, Théo Galy-Fajou, Matthäus Deutsch, Marius Kloft
This record contains the figures of the the ECML PKDD 2017 paper; Wenzel et al.: Bayesian Nonlinear Support Vector Machines for Big Data. All files are openly accessible in .pdf format.
accuracy_vs_inducing_points.pdf - average prediction error and training time as functions of the number of inducing points selected by two different methods with one standard deviation (using 10-fold cross validation).
accuracy_vs_time.pdf - prediction error as
function of training time for various methods discussed in the related paper.
autotuning_vs_gridsearch.pdf - average validation loss as function of the RBF kernel length-scale parameter q, computed by grid search and 10-fold cross validation. The red circle represents the hyperparameter found by our proposed automatic tuning approach.
Background
We propose a fast inference method for Bayesian nonlinear support vector machines that leverages stochastic variational inference and inducing points. Our experiments show that the proposed method is faster than competing Bayesian approaches and scales easily to millions of data points. It provides additional features over frequentist competitors such as accurate predictive uncertainty estimates and automatic hyperparameter search.