(Ph.D. in Computer Science, University of Milan, 2016)

Lecturer: Dr. Nicola Basilico

This course provides an introduction to multiagent systems by concentrating on modeling agents interactions by means of competitive games. The main objectives of this course are: conveying basic notions of game theoretical models, discussing in detail some of the algorithms for their resolution, and presenting some recent real-world applications. The course has 20 hours of class lectures and is worth 3 credits (CFU).(Course @ Phd School's website)

Announcements

  • For non-UNIMI students: when certifying the exam I can recognize additional hours for the final project if required by your PhD School's regulations — NB 2016/06/06 11:24
  • Course notes and other material presented in class have been completely uploaded — NB 2016/06/06 11:24
  • The calendar has been updated: the class of May 10th is postponed to May 25th — NB 2016/05/08 13:14
  • A doodle has been set up for rescheduling May 10th lecture — NB 2016/05/05 11:42
  • Course notes have been added for download — NB 2016/05/05 11:42
  • The calendar has been updated: the class of April 22th is postponed to May 10th — NB 2016/04/19 13:33
  • Anybody interested in attending this course should send an email to nicola.basilico@unimi.it — NB 2016/04/12 09:52

Class Schedule

Class sessions will be held from 9:00 to 12:15 (with breaks)

when where
April, 19th Meeting room 5 (“Auletta 5”) - dep. of Computer Science, Via Comelico 39 - Milano
April, 22th Meeting room 4 (“Auletta 4”) - dep. of Computer Science, Via Comelico 39 - Milano
April, 26th Meeting room 5 (“Auletta 5”) - dep. of Computer Science, Via Comelico 39 - Milano
May, 3th Meeting room 5 (“Auletta 5”) - dep. of Computer Science, Via Comelico 39 - Milano
May, 10th Meeting room 4 (“Auletta 4”) - dep. of Computer Science, Via Comelico 39 - Milano
May, 24th Meeting room 5 (“Auletta 5”) - dep. of Computer Science, Via Comelico 39 - Milano
May, 25th Meeting room 5 (“Auletta 5”) - dep. of Computer Science, Via Comelico 39 - Milano

Syllabus

  1. Introduction to Algorithmic Game Theory, self-interested agents, von Neumann-Morgenstern preferences and utilities, definition and examples of strategic form games, strategy profiles and expected utility (April 19th, 2016);
  2. Strategy profiles, strictly competitive games, solution concepts, Pareto efficiency, strict, weak and very weak dominance, dominant strategies, iterated removal of dominated actions (April 26th, 2016);
  3. Algorithms for dominance, Nash, Maxmin/Minmax strategies (May 3th, 2016);
  4. Nash and Maxmin, maxmin/minmax formulation and relations between the two; computing solution concepts: LP, linear complementarity (Lemke-Howson), Support Enumeration (Porter, Nudelman, Shoam), MIP-Nash (Sandholm, Gilpin, Conitzer) (May 24th, 2016);
  5. Correlated equilibrium, Leader-Follower equilibrium, applications: security games, software protection (May 25th, 2016).

Assignment

The required assignment promotes research thinking within the scope of the addressed topics. The course will provide basic notions that students are encouraged to apply to any problem (preferably but not necessarily in their area of expertise) they find interesting or significant, the objective being that of identifying (and, optionally, pursuing) possible research challenges. These are the main (not mutually exclusive) lines of work:

  • find a problem you are interested in and try to model it as game played by multiple agents; describe your formalization and discuss some benefits and research challenges that it would introduce;
  • select one the topics introduced in the course and briefly review the recent state-of-art on it or on some of its applications envisioning possible future directions of development;
  • develop one or more tools that can be useful for the understanding of the course topics or that can support some insights discussed in the previous lines.

The minimal deliverable for the course assignment is a short report of at least 3 pages (lncs LaTeX format). Cooperation and discussion between students is encouraged (especially if from different research backgrounds). Assignments can be done by parties of at most 3 students, the workload should be equally divided and the contribution of each one should be explicitly pointed out in the report.

Grading

Students that deliver the assignment will receive a grade on a scale from A to E. To successfully register the course, a grade of at least C is required.