Continual Learning Tasks

该数据集包含了一系列任务，其中数据和标签可能会在各个任务间发生变化，其设计旨在评估连续学习系统在稳定性、可塑性以及泛化能力方面的表现。此外，数据集还包含了多项指标，如当前任务准确率（cur-ACC）、上一任务准确率（last-ACC）、向后转移（BWT）、向前转移（FWT）以及一种新颖的指标——双重转移匹配指数（DMI）。该数据集的任务是评估连续学习各项指标的性能。

This dataset comprises a series of tasks, where the data and labels may vary across different tasks. It is designed to evaluate the performance of continual learning systems in terms of stability, plasticity, and generalization capability. In addition, the dataset includes multiple metrics, such as current task accuracy (cur-ACC), previous task accuracy (last-ACC), backward transfer (BWT), forward transfer (FWT), and a novel metric: the Dual Transfer Matching Index (DMI). The core task of this dataset is to assess the performance of continual learning based on these various metrics.