ABSTRACT
The goal of automatic classification is to predict at best the class y of an object x from some observations. A typical example is the spam filter of our mailbox, which predicts (more or less fairly) whether a mail is a spam or not. It is omnipresent in our daily life, by filtering the spams in our mailbox, reading automatically post-code on our postal letters, or recognizing faces on photos that we post on social networks. It is also very important in sciences, e.g., in medicine for early diagnosis of diseases from high-throughput data and in the industry, e.g., for detecting potential customers from their profiles.
