Thesis supervisor: Balázs Gerencsér
Location of studies (in Hungarian): ELTE Faculty of Science Department of Probability Theory and Statistics Abbreviation of location of studies: ELTE
Description of the research topic:
"The fundamental mathematical question is the asymptotics of distributed optimization on a graph, i.e., minimizing the sum_i f_i(x) where x needs to be agreed upon, but each (equal rank) vertex only knows its own f_i. There are some promising directions to further understand the convergence speed in certain scenarios. The investigation expands as we start to engage with applied motivations.
A crucial fundamental element of the process is simply averaging the vertices' initial values (still through a distributed algorithm).
There is a close connection with the mixing of Markov chains and the asymptotics of random matrix products."
Required language skills: English Further requirements: Probability Theory (possibly Dynamical Systems)