In case you don't think yourself have a strong background in probability theory, I would recommend the book by Ralf Herbrich "Learning Kernel Classifiers". This book seems hard to read in the beginning because of heavy mathematical notation. It is quite easy to follow when you drink some ice cold water and calm down. Especially noteworthy is the derivation of VC-dimension based bounds, which is the few book/papers I read that explain how those strange equations are obtained. In addition, the book "Kernel Methods for Pattern Analysis by Nello Cristianini is also very good and readable. All in all, Vapnik's books is the foundation.

Скачать книгу бесплатно (pdf, 29.82 Mb)

---

Читать «Statistical Learning Theory»