NavigatingMemoryConstructionbyGlobal Pseudo-TaskSimulationforContinualLearning

In the methods, they stick to a single static memory construction policy, which is prone to failing in the long task sequence.