Statistical and neural classifiers: an integrated approach to design
Statistical and neural classifiers: an integrated approach to design
Estimating a quality of decision function by empirical risk
MLDM'03 Proceedings of the 3rd international conference on Machine learning and data mining in pattern recognition
Hi-index | 0.00 |
Deciding functions statistical robustness problem is considered. The goal is to estimate misclassification probability for decision function by training sample. The paper contains results of investigation an empirical risk bias for decision tree classifier in comparison with a linear classifiers and with exact bias estimates for a discrete case. This allows to find out how far Vapnik–Chervonenkis risk estimates are off for considered decision function classes and to choose optimal complexity parameters for constructed decision functions. For heuristic algorithms those do not perform exhaustive search there was proposed a method for estimating an effective capacity.