A weighted combination of classifiers employing shared and distinct representations

This paper presents a theoretical framework for the combination of soft decisions generated by experts employing mixed (some shared and some distinct) object representations. By taking the confidence of the individuals experts into account, weighted benevolent fusion strategies are derived. This provides a basis for combining classifiers and illustrates that a substantial gain in performance can be achieved by using the opinions of multiple experts. These strategies are experimentally tested and their effectiveness is considered.

[1]  Kevin W. Bowyer,et al.  Combination of multiple classifiers using local accuracy estimates , 1996, Proceedings CVPR IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[2]  Adam Krzyżak,et al.  Methods of combining multiple classifiers and their applications to handwriting recognition , 1992, IEEE Trans. Syst. Man Cybern..

[3]  Kagan Tumer,et al.  Limits to performance gains in combined neural classifiers , 1995 .

[4]  Josef Kittler,et al.  Combining classifiers , 1996, Proceedings of 13th International Conference on Pattern Recognition.

[5]  Kagan Tumer,et al.  Analysis of decision boundaries in linearly combined neural classifiers , 1996, Pattern Recognit..

[6]  Josef Kittler,et al.  Strategies for combining classifiers employing shared and distinct pattern representations , 1997, Pattern Recognit. Lett..

[7]  J. Swets.  ROC analysis applied to the evaluation of medical imaging techniques. , 1979, Investigative radiology.

[8]  Ian W. Ricketts,et al.  The Mammographic Image Analysis Society digital mammogram database , 1994 .