Get Started
Research questionHow can adversarial online maximization of non-monotone DR-submodular functions achieve the best offline approximation factor?Adversarially changing objectives make it difficult to retain the approximation quality available when optimizing a fixed function offline. The difficulty is sharper for non-monotone objectives, where additional decisions can reduce value and feedback may be available only through an oracle.
Machine Learning
Research Paper
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Online Non-Monotone DR-Submodular Maximization Matching the Offline $0.401$ FactorThe setting is nonnegative non-monotone DR-submodular maximization over compact convex down-closed subsets of the unit cube. The stated guarantees use post-decision full-information value-oracle feedback that is conditionally unbiased and bounded, achieving a 0.401 approximation with sublinear approximate regret; a positive-anchor condition is additionally used for the one-point bandit result.research paper · Sep 2, 2026
Related questions
How can offline preference optimization identify which chosen–rejected pairs merit gradients without destabilizing reasoning-model training?When can reasoning LLMs reliably optimize numerical and semantically rich discrete search spaces in batches?How can functional bilevel optimization adapt online as learning objectives change over time?How can robotics optimization efficiently certify candidate solutions despite degenerate semidefinite relaxations?
Home
Topics
Search
Library