Distributed Optimization
and Games (DOG)
Giovanni Neglia
Communication Networks
 Many global optimization problems
 Avoid congestion, minimize energy consumption,
minimize file distribution time…
 But centralized optimization is unfeasible
 How to engineer distributed algorithms where each agent
contributes to solve the problem using only local
information?
Local interactions among agents in a
network have a global effect
Biology
Economy
Info
http://www-sop.inria.fr/members/Giovanni.Neglia/dog16/
(also look at 2015 course to know what’s coming)
[email protected]
Every lesson
 A short test (10-15 minutes) about the previous lesson
 Some specific examples/case studies
 take-home lessons
 Techniques/concepts to study similar problems
Resources (all available online!)
1) Book chapters
 Kelly&Yudovina, Stochastic networks
 Srinivas Shakkottai and R. Srikant, Network Optimization
and Control,
 David Easley and Jon Kleinberg, Networks, Crowds, and
Markets
 J. R. Marden and J.S. Shamma, Game Theory and
Distributed Control
2) Slides
3) Your notes
Evaluation
 in-class closed-book tests, top 5 out of 6 marks will count
for 15% of the final mark
 scribe notes, 1-person teams need to prepare only once,
2-person teams need to prepare twice (15% final mark)
• I wait for your teams and your dates
• Previous scribe notes available on github
 1 homework to be delivered at week 5 (15% final mark)
 Final exam (55% final mark)