M2I和I2M
收藏资源简介:
M2I和I2M是由AGH科技大学等机构创建的持续学习基准数据集,包含从MNIST到TinyImageNet(M2I)和反向(I2M)的六个不同图像分类数据集。这些数据集涵盖了从简单到复杂的任务,旨在评估持续学习模型在动态环境中的鲁棒性。数据集通过GitHub公开,支持研究社区进行严格的可重复性评估,特别强调模型在不断学习新任务的同时避免遗忘旧知识的能力。
M2I and I2M are continual learning benchmark datasets developed by institutions including AGH University of Science and Technology. They comprise six distinct image classification datasets, with M2I following the sequence from MNIST to TinyImageNet and I2M adopting the reverse order. These datasets cover tasks ranging from simple to complex, and are intended to assess the robustness of continual learning models in dynamic environments. Publicly available via GitHub, the datasets enable rigorous reproducibility assessments within the research community, with a particular focus on the ability of models to continuously learn new tasks while avoiding catastrophic forgetting of previously acquired knowledge.




