public class ClassificationFilter
extends java.lang.Object
The Classification Filter begins with n equal-sized disjoint subsets of the training set E (done with n-fold cross validation) and the empty output set A of detected noisy examples. The main loop is repeated for each training subset Ei. Ey is formed which includes all examples from E except those in Ei. Set Ey is used as the input for an arbitrary inductive learning algorithm that induces a hypothesis (a classifier) Hy. Those examples from Ei for which the hypothesis Hy does not give the correct classification are added to A as potentially noisy examples. Reference: 1999-Gamberger-ICML
Constructor and Description |
---|
ClassificationFilter()
It initializes the partitions from training set
|
Modifier and Type | Method and Description |
---|---|
void |
createDataset(java.lang.String out)
It apllies the changes to remove the noise
|
void |
run()
It initializes the partitions from training set
|