Open Set Learning

Open Set Learning (OSL) is a more challenging machine learning setting in the real world, aiming to handle cases where test samples contain new classes that did not appear during the training phase. Unlike traditional supervised learning, OSL not only requires the model to accurately classify samples of known categories but also to effectively identify and reject samples of unknown categories. This technology has significant application value in areas such as security monitoring and medical diagnosis, enhancing the robustness and generalization capabilities of systems.