Sequential Learning

In some applications, the environment may be so complex that it is unfeasible to choose a simple stochastic model and use classical statistical theory. A classic example is the spam detection which can be seen as a game between spammer and spam filters. Each trying to fool the other one. There is a necessity to take a robust approach by learning as ones goes along from experiences as more aspects of the problem are observed. This is the goal of online learning.

In online learning, data are acquired and treated on the fly; feedbacks are received and algorithms uploaded on the fly. This field has received a lot of attention recently because of the possible applications coming from internet. They include choosing which ads to display, repeated auctions, spam detection, experts/algorithm aggregation (and boosting), etc. The objectives of the course is to introduce and study the main concepts of online learning and design algorithms with theoretical analysis.

Prerequisite: probability theory (notion of random variables, convergence of random variables, conditional expectation).

Evaluation

This class is part of the Master 2 MVA. It will last 18 hours (3x6 lectures) + 2h for the exam.
Final grade: approximately 70% final exam, 30% homeworks (to implement some of the algorithms seen in class).

Exam

It will be on March 20th 2026, 10h-12h. A single two-sided sheet of handwritten notes (with any content) will be allowed for the exam. The exam will be divided into two separate parts (online convex optimization and stochastic bandits), each of which should be returned on a separate sheet. Please bring your own sheet of paper for the exam.

Homework

The homework is due by Friday, March 20, 2026. More information will be posted here. It can be done alone or in groups of two students. The report can be written in English or in French.

Reading material

• Prediction, learning, and games. N. Cesa-Bianchi and G. Lugosi, 2006.
• Regret Analysis of Stochastic and Nonstochastic Multi-armed Bandit Problems. S. Bubeck and N. Cesa-Bianchi, 2012.
• Bandit algorithms. T. Lattimore and Csaba Szepesvári, 2021.
• Introduction to online convex optimization. E. Hazan, 2016.
• A Modern Introduction to Online Learning. F. Orabona, 2019.

Contact: firstname.lastname@inria.fr