Then you ignored all the inaccuracies before that sentence, and all of the insights after it.

Have you ever trained an SVM? Because 10000 examples is a laughably small dataset. It's not a problem for modern implementations.

Linear SVMs can be trained on millions of examples.