The provided functions demonstrate a histogram-based measure for class separability, given the samples from two classes (binary classification problem). The proposed error classification estimation method is described in (B) and it is based on estimating the pdf of each class using histograms. The function that estimates the class seperability method is computeHistError(). Function theoreticalError() computes the theoretical error for two Gaussian distributed classes. Function testClassSeperability() calls the other two functions and displays the results for two Gaussian distributed functions. It has to be noted that computeHistError() can be used for any kind of class distribution, since it estimates the pdf of each class using the histogram method.
We can use computeHistError() in order to estimate the separabilty of a binary classification problem, given the training samples of the two classes.
-------------------------
Example
In order to execute the demo, call the testClassSeperability():
testClassSeperability(10000,1.0, 1.0, 3.0, 2.0, 1);
-------------------------------
Theodoros Giannakopoulos
http:/www.di.uoa.gr/~tyiannak
------------------------------- |
