Coordinate Descent with Bandit Sampling

被引:0
|
作者
Salehi, Farnood [1 ]
Thiran, Patrick [1 ]
Celis, L. Elisa [1 ]
机构
[1] Ecole Polytech Fed Lausanne, Sch Comp & Commun Sci, Lausanne, Switzerland
来源
ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 31 (NIPS 2018) | 2018年 / 31卷
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
Coordinate descent methods usually minimize a cost function by updating a random decision variable (corresponding to one coordinate) at a time. Ideally, we would update the decision variable that yields the largest decrease in the cost function. However, finding this coordinate would require checking all of them, which would effectively negate the improvement in computational tractability that coordinate descent is intended to afford. To address this, we propose a new adaptive method for selecting a coordinate. First, we find a lower bound on the amount the cost function decreases when a coordinate is updated. We then use a multi-armed bandit algorithm to learn which coordinates result in the largest lower bound by interleaving this learning with conventional coordinate descent updates except that the coordinate is selected proportionately to the expected decrease. We show that our approach improves the convergence of coordinate descent methods both theoretically and experimentally.
引用
收藏
页数:11
相关论文
共 50 条
  • [1] A RANDOMIZED COORDINATE DESCENT METHOD WITH VOLUME SAMPLING
    Rodomanov, Anton
    Kropotov, Dmitry
    SIAM JOURNAL ON OPTIMIZATION, 2020, 30 (03) : 1878 - 1904
  • [2] Faster Coordinate Descent via Adaptive Importance Sampling
    Perekrestenko, Dmytro
    Cevher, Volkan
    Jaggi, Martin
    ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 54, 2017, 54 : 869 - 877
  • [3] Accelerated Stochastic Block Coordinate Descent with Optimal Sampling
    Zhang, Aston
    Gu, Quanquan
    KDD'16: PROCEEDINGS OF THE 22ND ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING, 2016, : 2035 - 2044
  • [4] Distributed Proportional Stochastic Coordinate Descent With Social Sampling
    Ghassemi, Mohsen
    Sarwate, Anand D.
    2015 53RD ANNUAL ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING (ALLERTON), 2015, : 17 - 24
  • [5] Coordinate descent with arbitrary sampling I: algorithms and complexity
    Qu, Zheng
    Richtarik, Peter
    OPTIMIZATION METHODS & SOFTWARE, 2016, 31 (05): : 829 - 857
  • [6] Coordinate descent with arbitrary sampling II: expected separable overapproximation
    Qu, Zheng
    Richtarik, Peter
    OPTIMIZATION METHODS & SOFTWARE, 2016, 31 (05): : 858 - 884
  • [7] Accelerated Coordinate Descent with Arbitrary Sampling and Best Rates for Minibatches
    Hanzely, Filip
    Richtarik, Peter
    22ND INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 89, 2019, 89 : 304 - 312
  • [8] An Efficient Algorithm For Generalized Linear Bandit: Online Stochastic Gradient Descent and Thompson Sampling
    Ding, Qin
    Hsieh, Cho-Jui
    Sharpnack, James
    24TH INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS (AISTATS), 2021, 130
  • [9] A Unified Theory of SGD: Variance Reduction, Sampling, Quantization and Coordinate Descent
    Gorbunov, Eduard
    Hanzely, Filip
    Richtarik, Peter
    INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 108, 2020, 108 : 680 - 689
  • [10] When Cyclic Coordinate Descent Outperforms Randomized Coordinate Descent
    Gurbuzbalaban, Mert
    Ozdaglar, Asuman
    Parrilo, Pablo A.
    Vanli, N. Denizcan
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 30 (NIPS 2017), 2017, 30