Efficiently Learning Drifting Halfspaces with Massart Noise
Published in International Conference on Machine Learning (ICML 2026), 2026
We study learning a drifting concept in the presence of Massart noise for the class of margin-separable linear classifiers. We give a computationally efficient learner and characterize statistical-computational tradeoffs for this setting.
Recommended citation: Mingchen Ma, Guyang Cao, Jelena Diakonikolas, Ilias Diakonikolas. (2026). "Efficiently Learning Drifting Halfspaces with Massart Noise." Proceedings of the 43rd International Conference on Machine Learning (ICML 2026).
Download Paper