American Institute of Mathematical Sciences

Advanced
October  2012, 8(4): 1057-1069. doi: 10.3934/jimo.2012.8.1057

A proximal alternating direction method for $\ell_{2,1}$-norm least squares problem in multi-task feature learning

Yunhai Xiao 1, , Soon-Yi Wu 2, and  Bing-Sheng He 3,

1. 

Institute of Applied Mathematics, Henan University, Kaifeng 475004
2. 

National Center for Theoretical Sciences (South), National Cheng Kung University, Tainan 700, Taiwan
3. 

Department of Mathematics, Nanjing University, Nanjing 210093, China

Received  March 2011 Revised  January 2012 Published  September 2012

The joint feature selection problem arises in many fields including computer vision, text classification and biomedical informatics. Generally, recent results show that it can be realized by solving a $\ell_{2,1}$-norm involved minimization problem. However, solving the optimization problem is a challenging task due to the non-smoothness of the regularization term. In this paper, we reformulate the problem to an equivalent constrained minimization problem by introducing an auxiliary variable. We split the corresponding augmented Lagrange function and minimize the subproblem alternatively with one variable by fixing the other one. Moreover, we linearize the subproblem and add a proximal-point term when the closed-form solutions are not easily to derived. The convergence analysis and the relatedness with other algorithms are also given. Although the $\ell_{2,1}$-norm is mainly considered, we show that the $\ell_{\infty,1}$-norm penalized learning problem can also be readily solved in our framework. The reported experiments on simulated and real data sets show that the proposed method is effective and promising. The performance comparisons illustrate that the proposed algorithm is competitive with even performs little better than the state-of-the-art solver SLEP.
Keywords: proximal points, Multi-task feature learning, real data sets., alternating direction method, augmented Lagrangian function.
Mathematics Subject Classification: Primary: 90C30, 90C25; Secondary: 65K0.
Citation: Yunhai Xiao, Soon-Yi Wu, Bing-Sheng He. A proximal alternating direction method for $\ell_{2,1}$-norm least squares problem in multi-task feature learning. Journal of Industrial & Management Optimization, 2012, 8 (4) : 1057-1069. doi: 10.3934/jimo.2012.8.1057
References:
[1]

R. K. Ando and T. Zhang, A framework for learning predictive structures from multiple tasks and unlabeleddata,, Journal of Machine Learning Research, 6 (2005), 1817.   Google Scholar

[2]

A. Argyriou, T. Evgeniou and M. Pontil, Convex multi-convex feature learning,, Machine Learning, 73 (2008), 243.   Google Scholar

[3]

B. Bakker and T. Heskes, Task clustering and gating for Bayesian multi-task learning,, Journal of Machine Learning Research, 4 (2003), 83.   Google Scholar

[4]

S. Chen, D. Donoho and M. Saunders, Atomic decomposition by basis pursuit,, SIAM Journal on Scientific Computing, 20 (1999), 33.  doi: 10.1137/S1064827596304010.  Google Scholar

[5]

J. Duchi and Y. Singer, Efficient online and batch learning using forward backward splitting,, Journal of Machine Learning Research, 10 (2009), 2899.   Google Scholar

[6]

T. Evgeniou, C. A. Micchelli and M. Pontil, Learning multiple tasks with kernel methods,, Journal of Machine Learning Research, 6 (2005), 615.   Google Scholar

[7]

D. Gabay and B. Mercier, A dual algorithm for the solution of nonlinear variational problems via finite-element approximations,, Computers & Mathematics with Applications, 2 (1976), 17.   Google Scholar

[8]

R. Glowinski, "Numerical Methods for Nonlinear Variational Problems,", Springer, (1984).   Google Scholar

[9]

R. Glowinski and A. Marrocco, Sur l'approximation, par élémentsfinis d'ordre un, et la résolution, parpénalisation-dualité d'une classe de problèmes deDirichlet nonlinéaires,, Revue Francaise d'automatique, 2 (1975), 41.   Google Scholar

[10]

B. He, L. Z. Liao, D. Han and H. Yang, A new inexact alternating directions method for monotone variational inequalities,, Mathematical Programming, 92 (2002), 103.  doi: 10.1007/s101070100280.  Google Scholar

[11]

B. He, S. L. Wang and H. Yang, A modified variable-penalty alternating directions method for monotone variational inequalities,, Journal of Computational Mathematics, 21 (2003), 495.   Google Scholar

[12]

J. Liu, J. Chen and J. Ye, "Large-Scale Sparse Logistic Regression,", in, (2009).   Google Scholar

[13]

J. Liu, S. Ji and J. Ye, "Multi-Task Feather Learning Via Efficient $l_{2,1}$-norm Minimization,", in, (2009).   Google Scholar

[14]

M. Kowalski, Sparse regression using mixednorms,, Applied and Computational Harmonic Analysis, 27 (2009), 303.  doi: 10.1016/j.acha.2009.05.006.  Google Scholar

[15]

M. Kowalski, M. Szafranski and L. Ralaivola, "Multiple Indefinite Kernel Learning with Mixed Normregularization,", Proceedings of the 26th Annual International Conference on Machine Learning, (2009).   Google Scholar

[16]

A. Nemirovski, "Efficient Methods in Convex Programming,", Lecture Notes, (1994).   Google Scholar

[17]

Y. Nesterov, "Introductory Lectures on Convex Optimization: A Basic Course,", Kluwer Academic Publishers, (2003).   Google Scholar

[18]

Y. Nesterov, "Gradient Methods for Minimizing Composite Objective Function,", CORE report, (2007).   Google Scholar

[19]

F. Nie, H. Huang, X. Cai and C. Ding, "Efficient and Robust Feature Selection via Joint $l_{2,1}$-Normsminimization,", Neural Information Processing Systems Foundation, (2010).   Google Scholar

[20]

G. Obozinski, B. Taskar and M. I. Jordan, "Multi-Task Feature Selection,", Technical Report, (2006).   Google Scholar

[21]

Y. Saeys, I. Inza and P. Larranaga, A review of feature selection techniques in bioinformatics,, Bioinformatics, 23 (2007), 2507.  doi: 10.1093/bioinformatics/btm344.  Google Scholar

[22]

Y. Xiao, S.-Y. Wu and D.-H. Li, Splitting and linearizing augmented Lagrangian algorithm for subspace recovery from corrupted observations,, Adv. Comput. Math., (): 10444.   Google Scholar

[23]

T. Xiong, J. Bi, B. Rao and V. Cherkassky, "Probabilistic Joint Feature Selection for Multi-Task Learning,", in, (2006).   Google Scholar

[24]

M. H. Xu, Proximal alternating directions method for structured variational inequalities,, Journal of Optimization Theory and Applications, 134 (2007), 107.  doi: 10.1007/s10957-007-9192-2.  Google Scholar

[25]

J. Yang, Dynamic power price problem: An inverse variational inequality approach,, Journal of Industrial and Management Optimization, 4 (2008), 673.   Google Scholar

[26]

J. Yang and X. Yuan, Linearized augmented Lagrangian and alternating direction methods for nuclear norm minimization,, Math. Comput., ().  doi: 10.1090/S0025-5718-2012-02598-1.  Google Scholar

[27]

J. Yang and Y. Zhang, Alternating direction algorithms for $l_1$-problemsin compressive sensing,, SIAM Journal on Scientific Computing, 33 (2011), 250.  doi: 10.1137/090777761.  Google Scholar

[28]

J. Zhang, Z. Ghahramani and Y. Yang, Flexible latent variable models for multi-task learning,, Machine Learning, 73 (2008), 221.   Google Scholar

show all references

References:
[1]

R. K. Ando and T. Zhang, A framework for learning predictive structures from multiple tasks and unlabeleddata,, Journal of Machine Learning Research, 6 (2005), 1817.   Google Scholar

[2]

A. Argyriou, T. Evgeniou and M. Pontil, Convex multi-convex feature learning,, Machine Learning, 73 (2008), 243.   Google Scholar

[3]

B. Bakker and T. Heskes, Task clustering and gating for Bayesian multi-task learning,, Journal of Machine Learning Research, 4 (2003), 83.   Google Scholar

[4]

S. Chen, D. Donoho and M. Saunders, Atomic decomposition by basis pursuit,, SIAM Journal on Scientific Computing, 20 (1999), 33.  doi: 10.1137/S1064827596304010.  Google Scholar

[5]

J. Duchi and Y. Singer, Efficient online and batch learning using forward backward splitting,, Journal of Machine Learning Research, 10 (2009), 2899.   Google Scholar

[6]

T. Evgeniou, C. A. Micchelli and M. Pontil, Learning multiple tasks with kernel methods,, Journal of Machine Learning Research, 6 (2005), 615.   Google Scholar

[7]

D. Gabay and B. Mercier, A dual algorithm for the solution of nonlinear variational problems via finite-element approximations,, Computers & Mathematics with Applications, 2 (1976), 17.   Google Scholar

[8]

R. Glowinski, "Numerical Methods for Nonlinear Variational Problems,", Springer, (1984).   Google Scholar

[9]

R. Glowinski and A. Marrocco, Sur l'approximation, par élémentsfinis d'ordre un, et la résolution, parpénalisation-dualité d'une classe de problèmes deDirichlet nonlinéaires,, Revue Francaise d'automatique, 2 (1975), 41.   Google Scholar

[10]

B. He, L. Z. Liao, D. Han and H. Yang, A new inexact alternating directions method for monotone variational inequalities,, Mathematical Programming, 92 (2002), 103.  doi: 10.1007/s101070100280.  Google Scholar

[11]

B. He, S. L. Wang and H. Yang, A modified variable-penalty alternating directions method for monotone variational inequalities,, Journal of Computational Mathematics, 21 (2003), 495.   Google Scholar

[12]

J. Liu, J. Chen and J. Ye, "Large-Scale Sparse Logistic Regression,", in, (2009).   Google Scholar

[13]

J. Liu, S. Ji and J. Ye, "Multi-Task Feather Learning Via Efficient $l_{2,1}$-norm Minimization,", in, (2009).   Google Scholar

[14]

M. Kowalski, Sparse regression using mixednorms,, Applied and Computational Harmonic Analysis, 27 (2009), 303.  doi: 10.1016/j.acha.2009.05.006.  Google Scholar

[15]

M. Kowalski, M. Szafranski and L. Ralaivola, "Multiple Indefinite Kernel Learning with Mixed Normregularization,", Proceedings of the 26th Annual International Conference on Machine Learning, (2009).   Google Scholar

[16]

A. Nemirovski, "Efficient Methods in Convex Programming,", Lecture Notes, (1994).   Google Scholar

[17]

Y. Nesterov, "Introductory Lectures on Convex Optimization: A Basic Course,", Kluwer Academic Publishers, (2003).   Google Scholar

[18]

Y. Nesterov, "Gradient Methods for Minimizing Composite Objective Function,", CORE report, (2007).   Google Scholar

[19]

F. Nie, H. Huang, X. Cai and C. Ding, "Efficient and Robust Feature Selection via Joint $l_{2,1}$-Normsminimization,", Neural Information Processing Systems Foundation, (2010).   Google Scholar

[20]

G. Obozinski, B. Taskar and M. I. Jordan, "Multi-Task Feature Selection,", Technical Report, (2006).   Google Scholar

[21]

Y. Saeys, I. Inza and P. Larranaga, A review of feature selection techniques in bioinformatics,, Bioinformatics, 23 (2007), 2507.  doi: 10.1093/bioinformatics/btm344.  Google Scholar

[22]

Y. Xiao, S.-Y. Wu and D.-H. Li, Splitting and linearizing augmented Lagrangian algorithm for subspace recovery from corrupted observations,, Adv. Comput. Math., (): 10444.   Google Scholar

[23]

T. Xiong, J. Bi, B. Rao and V. Cherkassky, "Probabilistic Joint Feature Selection for Multi-Task Learning,", in, (2006).   Google Scholar

[24]

M. H. Xu, Proximal alternating directions method for structured variational inequalities,, Journal of Optimization Theory and Applications, 134 (2007), 107.  doi: 10.1007/s10957-007-9192-2.  Google Scholar

[25]

J. Yang, Dynamic power price problem: An inverse variational inequality approach,, Journal of Industrial and Management Optimization, 4 (2008), 673.   Google Scholar

[26]

J. Yang and X. Yuan, Linearized augmented Lagrangian and alternating direction methods for nuclear norm minimization,, Math. Comput., ().  doi: 10.1090/S0025-5718-2012-02598-1.  Google Scholar

[27]

J. Yang and Y. Zhang, Alternating direction algorithms for $l_1$-problemsin compressive sensing,, SIAM Journal on Scientific Computing, 33 (2011), 250.  doi: 10.1137/090777761.  Google Scholar

[28]

J. Zhang, Z. Ghahramani and Y. Yang, Flexible latent variable models for multi-task learning,, Machine Learning, 73 (2008), 221.   Google Scholar

[1]

Foxiang Liu, Lingling Xu, Yuehong Sun, Deren Han. A proximal alternating direction method for multi-block coupled convex optimization. Journal of Industrial & Management Optimization, 2019, 15 (2) : 723-737. doi: 10.3934/jimo.2018067
[2]

Xueyong Wang, Yiju Wang, Gang Wang. An accelerated augmented Lagrangian method for multi-criteria optimization problem. Journal of Industrial & Management Optimization, 2020, 16 (1) : 1-9. doi: 10.3934/jimo.2018136
[3]

Feng Ma, Jiansheng Shu, Yaxiong Li, Jian Wu. The dual step size of the alternating direction method can be larger than 1.618 when one function is strongly convex. Journal of Industrial & Management Optimization, 2020  doi: 10.3934/jimo.2020016
[4]

Fan Zhang, Guifa Teng, Mengmeng Gao, Shuai Zhang, Jingjing Zhang. Multi-machine and multi-task emergency allocation algorithm based on precedence rules. Discrete & Continuous Dynamical Systems - S, 2019, 12 (4&5) : 1501-1513. doi: 10.3934/dcdss.2019103
[5]

Russell E. Warren, Stanley J. Osher. Hyperspectral unmixing by the alternating direction method of multipliers. Inverse Problems & Imaging, 2015, 9 (3) : 917-933. doi: 10.3934/ipi.2015.9.917
[6]

Qian Liu, Xinmin Yang, Heung Wing Joseph Lee. On saddle points of a class of augmented lagrangian functions. Journal of Industrial & Management Optimization, 2007, 3 (4) : 693-700. doi: 10.3934/jimo.2007.3.693
[7]

Sohana Jahan. Supervised distance preserving projection using alternating direction method of multipliers. Journal of Industrial & Management Optimization, 2020, 16 (4) : 1783-1799. doi: 10.3934/jimo.2019029
[8]

Jie-Wen He, Chi-Chon Lei, Chen-Yang Shi, Seak-Weng Vong. An inexact alternating direction method of multipliers for a kind of nonlinear complementarity problems. Numerical Algebra, Control & Optimization, 2020  doi: 10.3934/naco.2020030
[9]

Yuan Shen, Wenxing Zhang, Bingsheng He. Relaxed augmented Lagrangian-based proximal point algorithms for convex optimization with linear constraints. Journal of Industrial & Management Optimization, 2014, 10 (3) : 743-759. doi: 10.3934/jimo.2014.10.743
[10]

Chunlin Wu, Juyong Zhang, Xue-Cheng Tai. Augmented Lagrangian method for total variation restoration with non-quadratic fidelity. Inverse Problems & Imaging, 2011, 5 (1) : 237-261. doi: 10.3934/ipi.2011.5.237
[11]

Wei Zhu, Xue-Cheng Tai, Tony Chan. Augmented Lagrangian method for a mean curvature based image denoising model. Inverse Problems & Imaging, 2013, 7 (4) : 1409-1432. doi: 10.3934/ipi.2013.7.1409
[12]

Li Jin, Hongying Huang. Differential equation method based on approximate augmented Lagrangian for nonlinear programming. Journal of Industrial & Management Optimization, 2020, 16 (5) : 2267-2281. doi: 10.3934/jimo.2019053
[13]

Bingsheng He, Xiaoming Yuan. Linearized alternating direction method of multipliers with Gaussian back substitution for separable convex programming. Numerical Algebra, Control & Optimization, 2013, 3 (2) : 247-260. doi: 10.3934/naco.2013.3.247
[14]

Yuan Shen, Lei Ji. Partial convolution for total variation deblurring and denoising by new linearized alternating direction method of multipliers with extension step. Journal of Industrial & Management Optimization, 2019, 15 (1) : 159-175. doi: 10.3934/jimo.2018037
[15]

Yue Lu, Ying-En Ge, Li-Wei Zhang. An alternating direction method for solving a class of inverse semi-definite quadratic programming problems. Journal of Industrial & Management Optimization, 2016, 12 (1) : 317-336. doi: 10.3934/jimo.2016.12.317
[16]

Zhongming Wu, Xingju Cai, Deren Han. Linearized block-wise alternating direction method of multipliers for multiple-block convex programming. Journal of Industrial & Management Optimization, 2018, 14 (3) : 833-855. doi: 10.3934/jimo.2017078
[17]

Dan Li, Li-Ping Pang, Fang-Fang Guo, Zun-Quan Xia. An alternating linearization method with inexact data for bilevel nonsmooth convex optimization. Journal of Industrial & Management Optimization, 2014, 10 (3) : 859-869. doi: 10.3934/jimo.2014.10.859
[18]

Xi-Hong Yan. A new convergence proof of augmented Lagrangian-based method with full Jacobian decomposition for structured variational inequalities. Numerical Algebra, Control & Optimization, 2016, 6 (1) : 45-54. doi: 10.3934/naco.2016.6.45
[19]

Anis Theljani, Ke Chen. An augmented lagrangian method for solving a new variational model based on gradients similarity measures and high order regulariation for multimodality registration. Inverse Problems & Imaging, 2019, 13 (2) : 309-335. doi: 10.3934/ipi.2019016
[20]

Wei Zhu. A numerical study of a mean curvature denoising model using a novel augmented Lagrangian method. Inverse Problems & Imaging, 2017, 11 (6) : 975-996. doi: 10.3934/ipi.2017045

2019 Impact Factor: 1.366

Tools

Metrics

  • PDF downloads (49)
  • HTML views (0)
  • Cited by (5)

Other articles
by authors

[Back to Top]

Copyright © 2020 American Institute of Mathematical Sciences