Techniques for Identifying Mislabeled Training Examples in ILP Classification Problems
Sofie Verbaeten, T Cardoen · Lirias · 2002
We consider the problem of noisy training examples, more precisely mislabeled training examples, in the context of ILP classification problems. We address this problem by pre-processing the training set, i.e. by identifying and removing outliers from the training set. We study a number of filtering techniques, some of which were proposed in the literature for attribute-value problems. We evaluate these techniques on a Bongard data set, which we artificially corrupt with different levels of classification noise.