Brief

Researchers have developed new algorithms for online packet scheduling with deadlines and partial feedback, aiming to minimize alpha-regret. The algorithms achieve an upper bound of O(sqrt(KT)) alpha-regret, matching the standard bandit setting. For instances with a deadline of at most one clock cycle away, a deterministic algorithm provides the tightest possible competitive ratio. Notably, when the number of distinct packet types is finite, these algorithms can surpass the established Phi competitive ratio barrier. AI