An emаil prоvider uses аn SVM tо clаssify messages as spam оr not spam. Some legitimate promotional emails look very similar to spam, and some spam messages are carefully written to look normal. Explain why allowing a few training errors might actually improve the classifier.