docNumPerformance

Centre for Vision, Speech & Signal Processing

Performance Evaluation Class Structure and Usage

NumPerformanceC is a base class which describes an interface for classification performance evaluation methods such as Hold-Out and Cross-Validation. Essentially, a classifier is provided with a set of training samples and the class evaluates the probability of misclassification with a call to the PMC member function. An error counting method must also be supplied which must be derived from NumErrorC.

Depending on which algorithm is used, the data set is split up into sections which are used to train and test the classifier and thereby measure its performance. The method of error counting is used to determine how much each individual classification error will contribute to the PMC.

Author: Robert Crida, Created: 12/6/1998, Generated by DocCreate: March 20, 2001