Journal of University of Chinese Academy of Sciences >
Improved classification model via MPEC
Received date: 2009-03-18
Revised date: 2009-04-20
Online published: 2009-09-15
In this paper, we provide an improved form of MPEC model for data classification first proposed in Ref.[1]. We use β likelihood estimation instead of maximum likelihood estimation to estimate the parameters of data's probability distribution function (modeled by Gaussian mixture models). Our new model can avoid the contingent of unboundedness of the maximum likelihood function and excessive sensitivity of the maximum likelihood estimator to outliers, showing more robustness. Then we use filterSQP method to solve our β likelihood MPEC model as nonlinear program. Efficiency of the model is shown by primal numerical tests.
Key words： MPEC; β likelihood estimation; filterSQP; robustness
DING Fei , YIN Hong-Xia . Improved classification model via MPEC[J]. Journal of University of Chinese Academy of Sciences, 2009 , 26(5) : 599 -608 . DOI: 10.7523/j.issn.2095-6134.2009.5.003
[1] Mangasarian O L. Linear and nonlinear separation of patterns by linear programming
[J]. Operations Research, 1965, 13(3):444-452.
[2] Mangasarian O L. Multisurface method of pattern separation
[J]. IEEE Transactions on Information Theory, 1968,IT-14(6):801-807.
[3] Vapnik V. The nature of statistical learning theory
[M]. New York: Springer, 1995.
[4] Olson T, Pang J S, Priebe C. A likelihood-MPEC approach to target classification
[J]. Mathematical Programming, 2003, 96:1-31.
[5] Priebe C E. Nonhomogeneity analysis using borrowed strength
[J]. Journal of the American Statistical Association, 1996, 91: 1497-1503.
[6] Luo Z Q, Pang J S, Ralph D. Mathematical programs with equilibrium constraints
[M]. Cambridge: Cambridge University Press, 1996.
[7] Scheel H, Scholtes S. Mathematical programs with equilibrium constraints: stationarity, optimality, and sensitivity
[J]. Mathematics of Operations Research, 2000, 25: 1-22.
[8] Scholtes S, Stuhr M. Exact penalization of methematical programs with comple-mentarity constraints
[J]. SIAM Journal on Control and Optimization, 1999, 37: 617-652.
[9] Anitescu M. On using the elasticmode in nonlinear programming approaches to mathematical programs with complementarity constraints
[J]. SIAM Journal on Optimization, 2005, 15: 1203-1236.
[10] Fletcher R, Leyffer S. Solving mathematical program with complementarity constraints as nonlinear programs
[J]. Optimization Methods and Software, 2004, 19(1): 15-40.
[11] Fletcher R, Leyffer S, Ralph D, et al. Local convergence of SQP methods for mathematical programs with eqllilibrillln constraints
[J]. SIAM Journal on Optimization, 2006, 17(1): 259-286.
[12] Fletcher R, Leyffer S. Nonlinear programming without a penalty fllnction
[J]. Mathematical Programming, 2002, 91: 239-270.
[13] Fujisawa H, Eguchi S. Robust estimation in the normal mixture model
[J]. Journal of Statistical Planning and Inference, 2006, 136: 3989-4011.
[14] Zhang Z W, Shi Y, Zhang P, et al. A rough set-based multiple criteria linear programming approach for classification
[J]. Lecture Notes in Computer Science, 2008, 5102: 476-485.
/
| 〈 |
|
〉 |