Learning Rates for Nonconvex Pairwise Learning

被引:2
|
作者
Li, Shaojie [1 ]
Liu, Yong [1 ]
机构
[1] Renmin Univ China, Gaoling Sch Artificial Intelligence, Beijing 100872, Peoples R China
基金
中国国家自然科学基金; 北京市自然科学基金;
关键词
Convergence; Stability analysis; Measurement; Training; Statistics; Sociology; Optimization; Generalization performance; learning rates; nonconvex optimization; pairwise learning; EMPIRICAL RISK; ALGORITHMS; STABILITY; RANKING; MINIMIZATION;
D O I
10.1109/TPAMI.2023.3259324
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Pairwise learning is receiving increasing attention since it covers many important machine learning tasks, e.g., metric learning, AUC maximization, and ranking. Investigating the generalization behavior of pairwise learning is thus of great significance. However, existing generalization analysis mainly focuses on the convex objective functions, leaving the nonconvex pairwise learning far less explored. Moreover, the current learning rates of pairwise learning are mostly of slower order. Motivated by these problems, we study the generalization performance of nonconvex pairwise learning and provide improved learning rates. Specifically, we develop different uniform convergence of gradients for pairwise learning under different assumptions, based on which we characterize empirical risk minimizer, gradient descent, and stochastic gradient descent. We first establish learning rates for these algorithms in a general nonconvex setting, where the analysis sheds insights on the trade-off between optimization and generalization and the role of early-stopping. We then derive faster learning rates of order O(1/n) for nonconvex pairwise learning with a gradient dominance curvature condition, where n is the sample size. Provided that the optimal population risk is small, we further improve the learning rates to O(1/n(2)), which, to the best of our knowledge, are the first O(1/n(2)) rates for pairwise learning.
引用
收藏
页码:9996 / 10011
页数:16
相关论文
共 50 条
  • [21] Stock Price Ranking by Learning Pairwise Preferences
    Engin Tas
    Ayca Hatice Atli
    Computational Economics, 2024, 63 : 513 - 528
  • [22] Stock Price Ranking by Learning Pairwise Preferences
    Tas, Engin
    Atli, Ayca Hatice
    COMPUTATIONAL ECONOMICS, 2024, 63 (02) : 513 - 528
  • [23] Refined bounds for online pairwise learning algorithms
    Chen, Xiaming
    Lei, Yunwen
    NEUROCOMPUTING, 2018, 275 : 2656 - 2665
  • [24] Modularizing Deep Learning via Pairwise Learning With Kernels
    Duan, Shiyu
    Yu, Shujian
    Principe, Jose C.
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2022, 33 (04) : 1441 - 1451
  • [25] An Optimal Transport Analysis on Generalization in Deep Learning
    Zhang, Jingwei
    Liu, Tongliang
    Tao, Dacheng
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2023, 34 (06) : 2842 - 2853
  • [26] Fair pairwise learning to rank
    Cerrato, Mattia
    Koeppel, Marius
    Segner, Alexander
    Esposito, Roberto
    Kramer, Stefan
    2020 IEEE 7TH INTERNATIONAL CONFERENCE ON DATA SCIENCE AND ADVANCED ANALYTICS (DSAA 2020), 2020, : 729 - 738
  • [27] Competitive Learning With Pairwise Constraints
    Covoes, Thiago F.
    Hruschka, Eduardo R.
    Ghosh, Joydeep
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2013, 24 (01) : 164 - 169
  • [28] Online Pairwise Learning Algorithms
    Ying, Yiming
    Zhou, Ding-Xuan
    NEURAL COMPUTATION, 2016, 28 (04) : 743 - 777
  • [29] Low-rank Tensor Learning with Nonconvex Overlapped Nuclear Norm Regularization
    Yao, Quanming
    Wang, Yaqing
    Han, Bo
    Kwok, James T.
    JOURNAL OF MACHINE LEARNING RESEARCH, 2022, 23
  • [30] Online pairwise learning algorithms with convex loss functions
    Lin, Junhong
    Lei, Yunwen
    Zhang, Bo
    Zhou, Ding-Xuan
    INFORMATION SCIENCES, 2017, 406 : 57 - 70