Distributionally Robust Optimization with Data Geometry

被引:0
作者
Liu, Jiashuo [1 ]
Wu, Jiayun [1 ]
Li, Bo [2 ]
Cui, Peng [1 ]
机构
[1] Tsinghua Univ, Dept Comp Sci Technol, Beijing, Peoples R China
[2] Tsinghua Univ, Sch Econ & Management, Beijing, Peoples R China
来源
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 35 (NEURIPS 2022) | 2022年
基金
国家重点研发计划; 中国国家自然科学基金;
关键词
DIMENSIONALITY REDUCTION; MODELS;
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
Distributionally Robust Optimization (DRO) serves as a robust alternative to empirical risk minimization (ERM), which optimizes the worst-case distribution in an uncertainty set typically specified by distance metrics including f-divergence and the Wasserstein distance. The metrics defined in the ostensible high dimensional space lead to exceedingly large uncertainty sets, resulting in the underperformance of most existing DRO methods. It has been well documented that high dimensional data approximately resides on low dimensional manifolds. In this work, to further constrain the uncertainty set, we incorporate data geometric properties into the design of distance metrics, obtaining our novel Geometric Wasserstein DRO (GDRO). Empowered by Gradient Flow, we derive a generically applicable approximate algorithm for the optimization of GDRO, and provide the bounded error rate of the approximation as well as the convergence rate of our algorithm. We also theoretically characterize the edge cases where certain existing DRO methods are the degeneracy of GDRO. Extensive experiments justify the superiority of our GDRO to existing DRO methods in multiple settings with strong distributional shifts, and confirm that the uncertainty set of GDRO adapts to data geometry.
引用
收藏
页数:13
相关论文
共 31 条
  • [1] [Anonymous], 2017, Certifying some distributional robustness with principled adversarial training
  • [2] Arjovsky M., 2019, ABS190702893 CORR
  • [3] Laplacian eigenmaps for dimensionality reduction and data representation
    Belkin, M
    Niyogi, P
    [J]. NEURAL COMPUTATION, 2003, 15 (06) : 1373 - 1396
  • [4] Data-driven robust optimization
    Bertsimas, Dimitris
    Gupta, Vishal
    Kallus, Nathan
    [J]. MATHEMATICAL PROGRAMMING, 2018, 167 (02) : 235 - 292
  • [5] Chen RD, 2018, J MACH LEARN RES, V19
  • [6] The Discrete Moment Problem with Nonconvex Shape Constraints
    Chen, Xi
    He, Simai
    Jiang, Bo
    Ryan, Christopher Thomas
    Zhang, Teng
    [J]. OPERATIONS RESEARCH, 2021, 69 (01) : 279 - 296
  • [7] Chow Shui-Nee, 2017, ENTROPY DISSIPATION
  • [8] Differential abundance testing on single-cell data using k-nearest neighbor graphs
    Dann, Emma
    Henderson, Neil C.
    Teichmann, Sarah A.
    Morgan, Michael D.
    Marioni, John C.
    [J]. NATURE BIOTECHNOLOGY, 2022, 40 (02) : 245 - +
  • [9] Distributionally Robust Optimization Under Moment Uncertainty with Application to Data-Driven Problems
    Delage, Erick
    Ye, Yinyu
    [J]. OPERATIONS RESEARCH, 2010, 58 (03) : 595 - 612
  • [10] Dong W., 2011, WWW 11, P577, DOI [10.1145/1963405.1963487, DOI 10.1145/1963405.1963487]