Coverage-Maximizing Multinomial Subset Routing under Operational Constraints
2026-08-17 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors introduce a new way for a computer program to pick which experts to listen to by randomly sampling from a list, rather than always choosing the same group. The program only gets feedback from the best expert it picked each time, which is different from older methods that add up rewards from all chosen experts. They also make sure the program follows certain rules over time, even though it only sees limited feedback. The authors created an algorithm to handle this and showed it works well both in theory and with real data.
online routingmultinomial policybandit feedbacksubset selectioncombinatorial banditsonline mirror descentBlackwell's Approachabilityregret analysisoperational constraintscrowdsourcing
Authors
Quan Zhou, Yiyan Huang
Abstract
We introduce Multinomial Subset Routing (MSR), a new online routing framework over $K$ experts in which the learner keeps a multinomial routing policy instead of a deterministic subset of experts. At each round, the learner samples $M$ experts i.i.d. from the multinomial policy, and the resulting set of distinct sampled experts forms the routed subset. The reward depends only on the best-performing expert(s) in the routed subset. This reward structure arises naturally in routing across specialized models but is not captured by standard combinatorial bandits or subset-selection methods, which optimize deterministic subsets and typically assume additive rewards. We require the selection to satisfy several long-term, two-sided operational constraints under bandit feedback, observing only the winner's reward each round. We propose OMD-Approachability, combining online mirror descent with Blackwell's Approachability, and prove it achieves $O(1/\sqrt{T})$ regret in both reward and constraint violation. We ground the framework in practical application domains and validate it empirically on a real-world crowdsourcing dataset.