Online chairman assignment with polylogarithmic recourse

Obtain a constant discrepancy bound for the online chairman assignment problem under polylogarithmic recourse, where columns arrive online and a limited number of reassignment operations per arrival are allowed.

Background

In the online chairman assignment without recourse, the optimal discrepancy is Θ(log m) against adaptive adversaries (achieved by a simple greedy algorithm). Recent work studies the online carpooling problem with recourse, but analogous guarantees for the chairman assignment with recourse are unresolved.

A constant discrepancy bound with polylog recourse would significantly strengthen online discrepancy minimization for assignment problems.