Improve Speed of SVM Regression

I'm using regression to predict some data. The data set consists of 7 predictors with 26195 observations. If I train a regression model using decision trees, ensemble trees or linear regression, Matlab generates the model quickly. But if a try to train a support vector machine (SVM) regression model for any kernel, it takes forever. Ant tips on how to improve the speed?

[emmanuel adewumi]

I am also facing this challenge. Is there a way of improving the speed of any trained model in SVM?

[Michael Croucher]

Could you provide an example please? Something we could run ourselves?