Multi-Task Least-Squares Support Vector Regression Machines and Their Applications in NIR Spectral Analysis
XU Shuo1, QIAO Xiao-dong1, ZHU Li-jun1, AN Xin2, ZHANG Lu-da3*
1. Information Technology Supporting Center, Institute of Science and Technology Information of China, Beijing 100038, China 2. School of International Trade and Economics, University of International Business and Economics, Beijing 100029, China 3. College of Science, China Agricultural University, Beijing 100193, China
Abstract：In near infrared spectral quantitative analysis, many models consider separately each component when modeling sample composition content, disregarding the underlying relatedness among sample compositions. To address this problem, the present paper views modeling each sample composition content as a task, thus one can transform the problem that models simultaneously analyze all sample compositions’ contents to a multi-task learning problem. On the basis of the LS-SVR, a multi-task LS-SVR (MTLS-SVR) model is proposed. Furthermore, an efficient large-scale algorithm is given. The broomcorn samples are taken as experimental material, and corresponding quantitative analysis models are constructed for three sample composition contents (protein, lysine and starch) with LS-SVR, PLS, multiple dependent variables LS-SVR (MLS-SVR) and MTLS-SVR. For the MTLS-SVR model, the average relative errors between actual values and predicted ones for the three sample compositions contents are 1.52%, 3.04% and 1.01%, respectively, and the correlation coefficients are 0.993 1, 0.894 0 and 0.940 6, respectively. Experimental results show MTLS-SVR model outperforms significantly the three others, which verifies the feasibility and efficiency of the MTLS-SVR model.
[1] YAN Yan-lu, ZHAO Long-lian, HAN Dong-hai，et al(严衍禄，赵龙莲，韩东海, 等). Foundation of Near Infrared Spectral Analysis and its Applications(近红外光谱分析基础与应用). Beijing: China Light Industry Press(北京：中国轻工业出版社), 2005. [2] Abdi H. Partial Least Squares (PLS) Regression. Encyclopedia for Research Methods for the Social Sciences, Lewis-Beck M, Bryman A, Futing T, eds. Sage, Thousand Oaks, CA, pp.792. [3] Vapnik V N. The Nature of Statistical Learning Theory, 2nd Edition. New York: Springer Verlag, 1999. [4] Suykens J A K, Gestel T V, Brabanter J D，et al. Least Squares Support Vector Machines. Singapore：World Scientific Pub. Co., 2002. [5] Bakker B，Heskes T. Journal of Machine Learning Research, 2003, 4(May): 83. [6] Heskes T. Empirical Bayes for Learning to Learn. Proceedings of the 17th International Conference on Machine Learning (ICML), San Francisco, CA, USA, 2000. 367. [7] Evgeniou T，Pontil M. Regularized Multi-Task Learning. Proceedings of the 10th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), Seattle, WA, USA, 2004. 109. [8] Golub G H，Van Loan C F. Matrix Computations, 3rd Edition. Baltimore and London：Johns Hopkins University Press, 1996. [9] Press W H, Teukolsky S A, Vetterling W T，et al. Numerical Recipes in C: The Art of Scientific Computing. New York：Cambridge University Press, 1992. [10] Saad Y. Iterative Methods for Sparse Linear Systems, 2nd Edition. Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA, USA, 2003. [11] Hamers B, Suykens J A K，Moor B D. A Comparison of Iterative Methods for Least Squares Vector Machine Classifiers. Internal Report 01-110, 2001. ESAT-SISTA, K.U. Leuven, Belgium. [12] Xu S, Ma F J，Tao L. Learn from the Information Contained in the False Splice Sites as well as in the True Splice Sites using SVM. Proceedings of the International Conference on Intelligent Systems and Knowledge Engineering (ISKE), Chengdu, China, 2007. 1360. [13] Rosipal R，Trejo L J. Journal of Machine Learning Research, 2001, 2(Dec): 97.
