Timezone: »

Bandit Learning with Positive Externalities
Virag Shah · Jose Blanchet · Ramesh Johari

Thu Dec 06 02:00 PM -- 04:00 PM (PST) @ Room 517 AB #158

In many platforms, user arrivals exhibit a self-reinforcing behavior: future user arrivals are likely to have preferences similar to users who were satisfied in the past. In other words, arrivals exhibit {\em positive externalities}. We study multiarmed bandit (MAB) problems with positive externalities. We show that the self-reinforcing preferences may lead standard benchmark algorithms such as UCB to exhibit linear regret. We develop a new algorithm, Balanced Exploration (BE), which explores arms carefully to avoid suboptimal convergence of arrivals before sufficient evidence is gathered. We also introduce an adaptive variant of BE which successively eliminates suboptimal arms. We analyze their asymptotic regret, and establish optimality by showing that no algorithm can perform better.

Author Information

Virag Shah (Stanford)
Jose Blanchet (Stanford University)
Ramesh Johari (Stanford University)

More from the Same Authors