Neural Network Optimization for Combinations in Identification Systems

Authors:
Sergey Tulyakov;Venu Govindaraju
Affiliations:
Center for Unified Biometrics and Sensors, University at Buffalo, Buffalo, USA;Center for Unified Biometrics and Sensors, University at Buffalo, Buffalo, USA
Venue:
MCS '09 Proceedings of the 8th International Workshop on Multiple Classifier Systems
Year:
2009

Citing 6
Cited 0

On Combining Classifiers

IEEE Transactions on Pattern Analysis and Machine Intelligence
Neural Networks for Pattern Recognition

Neural Networks for Pattern Recognition
Pattern Recognition and Neural Networks

Pattern Recognition and Neural Networks
A Bayesian Committee Machine

Neural Computation
Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning)

Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning)
Kernel carpentry for online regression using randomly varying coefficient model

IJCAI'07 Proceedings of the 20th international joint conference on Artifical intelligence

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper we investigate the construction of combination functions in identification systems. In contrast to verification systems, the optimal combination functions for identification systems are not known. In this paper we represent the combination function by means of a neural network and explore different methods of its training, so that the identification system performance is optimized. The modifications are based on the principle of utilizing best impostors from each training identification trial. The experiments are performed on score sets of biometric matchers and handwritten word recognizers. The proposed combination methods are able to outperform the likelihood ratio, which is optimal combination method for verification system, as well as, weighted sum combination method optimized for best performance in identification systems.