Comparing decision boundary curvature
Hui Zhu, Jianhua Huang, Xianglong Tang · Proceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004. · 2004
Classifier decision boundary is important for classification. By studying the properties of decision boundary, we can predict the classifier's performance. An algorithm of comparing classifiers' decision boundary curvature is introduced in this paper. Geometrical methods are used to carry out the comparison. This comparison can be used to measure the difference between classifiers which is useful for designing multiple classifier systems. The effectiveness of this comparison is confirmed by experiments.