Multivariate Polynomials Estimation Based on GradientBoost in Multimodal Biometrics
One of the traditional criteria to estimate the value of coefficients of multivariate polynomials in regression applications is MSE, which is known as OWM in classifier combination literature. In this paper, we address the use of GradientBoost algorithm to estimate coefficients of multivariate polynomials for score fusion level in multimodal biometric systems. Our experiments on NIST-bssr1 score database showed an improvement in verification accuracy and also reduction of number of coefficients, which increased the memory efficiency. In addition, we examined combination of OWM, and GradientBoost which showed better ROC performance and lower model order compared to OWM alone.
KeywordsMultimodal Biometrics Multivariate Polynomials GradientBoost Classifiers Fusion
Unable to display preview. Download preview PDF.
- 2.Fierrez-Aguilar, J.: Adapted Fusion Schemes for Multimodal Biometric Authentication. Phd Thesis, Universidad Politecnica de Madrid (2006)Google Scholar
- 6.Collins, M., Schapire, R.E., Singer, Y.: Logistic Regression, AdaBoost and Bregman Distances. In: COLT, pp. 158–169 (2000)Google Scholar
- 8.Friedman, J.H.: Greedy Function Approximation: a GradientBoost Machine. Technical report, Department of Statistics, Stanford University (1999)Google Scholar
- 10.National Institute of Standards and Technology. NIST Biometric Scores Set - release 1 (2004), http://www.tl.nist.gov/iad/894.03/biometricscores