Label noise is ever-present in machine learning practice. Allegro datasets are no exception. We compared 7 methods for training classifiers robust to label noise. All of them improved the model’s performance on noisy datasets. Some of the methods decreased the model’s performance in the absence of label noise.
Senior Research Engineer in the Machine Learning Research team, working on NLP in the e-commerce domain. Kalina Kobus obtained her PhD in Machine Learning from Poznan University of Technology. Her research interests include large-scale text classification, deep metric learning and the robustness of classifiers against mislabeled training data.
