Differentiable Submodular Maximization

被引:0
作者
Tschiatschek, Sebastian [1 ]
Sahin, Aytunc [2 ]
Krause, Andreas [2 ]
机构
[1] Microsoft Res Cambridge, Cambridge, England
[2] Swiss Fed Inst Technol, Zurich, Switzerland
来源
PROCEEDINGS OF THE TWENTY-SEVENTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE | 2018年
基金
瑞士国家科学基金会; 欧洲研究理事会;
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We consider learning of submodular functions from data. These functions are important in machine learning and have a wide range of applications, e.g. data summarization, feature selection and active learning. Despite their combinatorial nature, submodular functions can be maximized approximately with strong theoretical guarantees in polynomial time. Typically, learning the submodular function and optimization of that function are treated separately, i.e. the function is first learned using a proxy objective and subsequently maximized. In contrast, we show how to perform learning and optimization jointly. By interpreting the output of greedy maximization algorithms as distributions over sequences of items and smoothening these distributions, we obtain a differentiable objective. In this way, we can differentiate through the maximization algorithms and optimize the model to work well with the optimization algorithm. We theoretically characterize the error made by our approach, yielding insights into the tradeoff of smoothness and accuracy. We demonstrate the effectiveness of our approach for jointly learning and optimizing on synthetic maximum cut data, and on real world applications such as product recommendation and image collection summarization.
引用
收藏
页码:2731 / 2738
页数:8
相关论文
共 50 条
[31]   Robust Maximization of Correlated Submodular Functions [J].
Hou, Qiqiang ;
Clark, Andrew .
2019 IEEE 58TH CONFERENCE ON DECISION AND CONTROL (CDC), 2019, :7177-7183
[32]   Resilient Monotone Submodular Function Maximization [J].
Tzoumas, Vasileios ;
Gatsis, Konstantinos ;
Jadbabaie, Ali ;
Pappas, George J. .
2017 IEEE 56TH ANNUAL CONFERENCE ON DECISION AND CONTROL (CDC), 2017,
[33]   Sequence submodular maximization meets streaming [J].
Yang, Ruiqi ;
Xu, Dachuan ;
Guo, Longkun ;
Zhang, Dongmei .
JOURNAL OF COMBINATORIAL OPTIMIZATION, 2021, 41 (01) :43-55
[34]   On Distributed Submodular Maximization with Limited Information [J].
Gharesifard, Bahman ;
Smith, Stephen L. .
2016 AMERICAN CONTROL CONFERENCE (ACC), 2016, :1048-1053
[35]   Streaming Algorithms for Submodular Function Maximization [J].
Chekuri, Chandra ;
Gupta, Shalmoli ;
Quanrud, Kent .
AUTOMATA, LANGUAGES, AND PROGRAMMING, PT I, 2015, 9134 :318-330
[36]   Federated Submodular Maximization With Differential Privacy [J].
Wang, Yanhao ;
Zhou, Tianchen ;
Chen, Cen ;
Wang, Yinggui .
IEEE INTERNET OF THINGS JOURNAL, 2024, 11 (02) :1827-1839
[37]   Streaming Submodular Maximization with the Chance Constraint [J].
Gong, Shufang ;
Liu, Bin ;
Fang, Qizhi .
FRONTIERS OF ALGORITHMIC WISDOM, IJTCS-FAW 2022, 2022, 13461 :129-140
[38]   Robust monotone submodular function maximization [J].
James B. Orlin ;
Andreas S. Schulz ;
Rajan Udwani .
Mathematical Programming, 2018, 172 :505-537
[39]   SUBMODULAR MAXIMIZATION WITH UNCERTAIN KNAPSACK CAPACITY [J].
Kawase, Yasushi ;
Sumita, Hanna ;
Fukunagm, Takuro .
SIAM JOURNAL ON DISCRETE MATHEMATICS, 2019, 33 (03) :1121-1145
[40]   Optimal Region Search with Submodular Maximization [J].
Chen, Xuefeng ;
Cao, Xin ;
Zeng, Yifeng ;
Fang, Yixiang ;
Yao, Bin .
PROCEEDINGS OF THE TWENTY-NINTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2020, :1216-1222