April 28th14:30 , R2014 Digiteo Shannon (660) (see location):
Title: Optimal Sample Size in Multi-Phase Learning
Motivated by practical applications, chiefly clinical trials, we study the regret achievable for stochastic multi-armed bandits under the constraint that the employed policy must function in a small number of phases. Our results show that a very small number of phases gives already close to minimax optimal regret bounds and we also evaluate the number of trials in each phase.
Contact: cyril.furtlehner at inria.fr