# American Institute of Mathematical Sciences

• Previous Article
Optimality conditions for $E$-differentiable vector optimization problems with the multiple interval-valued objective function
• JIMO Home
• This Issue
• Next Article
Two nonparametric approaches to mean absolute deviation portfolio selection model
doi: 10.3934/jimo.2019052

## Improved SVRG for finite sum structure optimization with application to binary classification

 1 School of Computer Science and Technology, Anhui University of Technology, Maanshan 243032, China 2 School of Sciences, Hangzhou Dianzi University, Hangzhou 310018, China

* Corresponding author: Wei Xue (E-mail: cswxue@ahut.edu.cn)

Received  May 2018 Revised  November 2018 Published  May 2019

This paper looks at a stochastic variance reduced gradient (SVRG) method for minimizing the sum of a finite number of smooth convex functions, which has been involved widely in the field of machine learning and data mining. Inspired by the excellent performance of two-point stepsize gradient method in batch learning, in this paper we present an improved SVRG algorithm, named stochastic two-point stepsize gradient method. Under some mild conditions, the proposed method achieves a linear convergence rate $O(\rho^k)$ for smooth and strongly convex functions, where $\rho\in(0.68, 1)$. Simulation experiments on several benchmark data sets are reported to demonstrate the performance of the proposed method.

Citation: Guangmei Shao, Wei Xue, Gaohang Yu, Xiao Zheng. Improved SVRG for finite sum structure optimization with application to binary classification. Journal of Industrial & Management Optimization, doi: 10.3934/jimo.2019052
##### References:

show all references

##### References:
Objective loss and test classification accuracy versus epochs on data set "a9a" (best viewed in color)
Objective loss and test classification accuracy versus epochs on data set "w8a"
Evolutions of objective loss and test classification accuracy w.r.t. epochs for Eq. (14) on data set "a9a"
Evolutions of objective loss and test classification accuracy w.r.t. epochs for Eq. (14) on data set "w8a"
Evolutions of objective loss and test classification accuracy w.r.t. epochs for Eq. (14) on data set "a1a"
Evolutions of objective loss and test classification accuracy w.r.t. epochs for Eq. (14) on data set "w1a"
Evolutions of objective loss and test classification accuracy versus epochs on data set "a9a" with different $\eta_0$
Evolutions of objective loss and test classification accuracy versus epochs on data set "w8a" with different $\eta_0$
Comparison results between SVM and STSG on test classification accuracy on four test data sets
Details of data sets
 Data set $\sharp$ of training samples $\sharp$ of test samples $\sharp$ of dimension a1a 1605 30956 123 a9a 32561 16281 123 w1a 2477 47272 300 w8a 49749 14951 300
 Data set $\sharp$ of training samples $\sharp$ of test samples $\sharp$ of dimension a1a 1605 30956 123 a9a 32561 16281 123 w1a 2477 47272 300 w8a 49749 14951 300
 [1] Ning Zhang, Qiang Wu. Online learning for supervised dimension reduction. Mathematical Foundations of Computing, 2019, 2 (2) : 95-106. doi: 10.3934/mfc.2019008 [2] Wen-Chiao Cheng. Two-point pre-image entropy. Discrete & Continuous Dynamical Systems - A, 2007, 17 (1) : 107-119. doi: 10.3934/dcds.2007.17.107 [3] Feliz Minhós, A. I. Santos. Higher order two-point boundary value problems with asymmetric growth. Discrete & Continuous Dynamical Systems - S, 2008, 1 (1) : 127-137. doi: 10.3934/dcdss.2008.1.127 [4] Wenming Zou. Multiple solutions results for two-point boundary value problem with resonance. Discrete & Continuous Dynamical Systems - A, 1998, 4 (3) : 485-496. doi: 10.3934/dcds.1998.4.485 [5] Frédéric Legoll, William Minvielle. Variance reduction using antithetic variables for a nonlinear convex stochastic homogenization problem. Discrete & Continuous Dynamical Systems - S, 2015, 8 (1) : 1-27. doi: 10.3934/dcdss.2015.8.1 [6] Jerry L. Bona, Hongqiu Chen, Shu-Ming Sun, Bing-Yu Zhang. Comparison of quarter-plane and two-point boundary value problems: The KdV-equation. Discrete & Continuous Dynamical Systems - B, 2007, 7 (3) : 465-495. doi: 10.3934/dcdsb.2007.7.465 [7] Xiao-Yu Zhang, Qing Fang. A sixth order numerical method for a class of nonlinear two-point boundary value problems. Numerical Algebra, Control & Optimization, 2012, 2 (1) : 31-43. doi: 10.3934/naco.2012.2.31 [8] Shao-Yuan Huang, Shin-Hwa Wang. On S-shaped bifurcation curves for a two-point boundary value problem arising in a theory of thermal explosion. Discrete & Continuous Dynamical Systems - A, 2015, 35 (10) : 4839-4858. doi: 10.3934/dcds.2015.35.4839 [9] Jerry Bona, Hongqiu Chen, Shu Ming Sun, B.-Y. Zhang. Comparison of quarter-plane and two-point boundary value problems: the BBM-equation. Discrete & Continuous Dynamical Systems - A, 2005, 13 (4) : 921-940. doi: 10.3934/dcds.2005.13.921 [10] Marcel Lesieur. Two-point closure based large-eddy simulations in turbulence, Part 1: Isotropic turbulence. Discrete & Continuous Dynamical Systems - S, 2011, 4 (1) : 155-168. doi: 10.3934/dcdss.2011.4.155 [11] Marcel Lesieur. Two-point closure based large-eddy simulations in turbulence. Part 2: Inhomogeneous cases. Discrete & Continuous Dynamical Systems - A, 2010, 28 (1) : 227-241. doi: 10.3934/dcds.2010.28.227 [12] Cai-Tong Yue, Jing Liang, Bo-Fei Lang, Bo-Yang Qu. Two-hidden-layer extreme learning machine based wrist vein recognition system. Big Data & Information Analytics, 2017, 2 (1) : 59-68. doi: 10.3934/bdia.2017008 [13] Shuhua Wang, Zhenlong Chen, Baohuai Sheng. Convergence of online pairwise regression learning with quadratic loss. Communications on Pure & Applied Analysis, 2020, 19 (8) : 4023-4054. doi: 10.3934/cpaa.2020178 [14] Émilie Chouzenoux, Henri Gérard, Jean-Christophe Pesquet. General risk measures for robust machine learning. Foundations of Data Science, 2019, 1 (3) : 249-269. doi: 10.3934/fods.2019011 [15] Aude Hofleitner, Tarek Rabbani, Mohammad Rafiee, Laurent El Ghaoui, Alex Bayen. Learning and estimation applications of an online homotopy algorithm for a generalization of the LASSO. Discrete & Continuous Dynamical Systems - S, 2014, 7 (3) : 503-523. doi: 10.3934/dcdss.2014.7.503 [16] Roberto C. Alamino, Nestor Caticha. Bayesian online algorithms for learning in discrete hidden Markov models. Discrete & Continuous Dynamical Systems - B, 2008, 9 (1) : 1-10. doi: 10.3934/dcdsb.2008.9.1 [17] Joon Kwon, Panayotis Mertikopoulos. A continuous-time approach to online optimization. Journal of Dynamics & Games, 2017, 4 (2) : 125-148. doi: 10.3934/jdg.2017008 [18] René Henrion, Christian Küchler, Werner Römisch. Discrepancy distances and scenario reduction in two-stage stochastic mixed-integer programming. Journal of Industrial & Management Optimization, 2008, 4 (2) : 363-384. doi: 10.3934/jimo.2008.4.363 [19] Jiang Xie, Junfu Xu, Celine Nie, Qing Nie. Machine learning of swimming data via wisdom of crowd and regression analysis. Mathematical Biosciences & Engineering, 2017, 14 (2) : 511-527. doi: 10.3934/mbe.2017031 [20] Mingbao Cheng, Shuxian Xiao, Guosheng Liu. Single-machine rescheduling problems with learning effect under disruptions. Journal of Industrial & Management Optimization, 2018, 14 (3) : 967-980. doi: 10.3934/jimo.2017085

2018 Impact Factor: 1.025

## Tools

Article outline

Figures and Tables