核心概念
Proposing unified projection-free algorithms for adversarial continuous DR-submodular optimization, achieving state-of-the-art results in various scenarios.
摘要
The paper introduces novel algorithms for online continuous adversarial DR-submodular optimization, addressing monotone and non-monotone functions, different feedback types, and convex feasible regions. The proposed algorithms outperform existing methods in terms of regret bounds and query complexities. By utilizing meta-actions, random permutations, and a smoothing trick, the algorithms achieve efficient optimization with fewer queries. Experimental results demonstrate superior performance compared to baseline methods.
統計資料
For every problem considered in the non-monotone setting, the proposed algorithms are either the first with proven sub-linear α-regret bounds or have better α-regret bounds than the state of the art.
In the monotone setting, the proposed approach gives state-of-the-art sub-linear α-regret bounds among projection-free algorithms in 7 of the 8 considered cases while matching the result of the remaining case.
引述
"Our algorithm’s regret bounds dominate the state of the art."
"The key contributions include a unified framework for Frank-Wolfe type algorithms for adversarial continuous DR-submodular optimization."