SVMs are a nice technique, but they’re slow (O(N³)) and often give worse classification results than techniques such as radial basis function networks or even bagged / boosted kNN, Bayes, and/or decision tree models. So why are they so popular?