Ensemble noisy label detection on MNIST
István Fazekas, Attila Barta, László Fórián · Az Eszterházy Károly Tanárképző Főiskola tudományos közleményei. Tanulmányok a matematikai tudományok köréből/Az Eszterházy Károly Főiskola tudományos közleményei. Tanulmányok a matematikai tudományok köréből/Annales mathematicae et informaticae · 2021
In this paper machine learning methods are studied for classification data containing some misleading items. We use ensembles of known noise correction methods for preprocessing the training set. Preprocessing can be either relabeling or deleting items detected to have noisy labels. After preprocessing, usual convolutional networks are applied to the data. With preprocessing, the performance of very accurate convolutional networks can be further improved.