1 paper
Shixiong Xu, Gaofeng Meng, Xing Nie +3
We observe a high level of imbalance in the accuracy of different classes in the same old task for the first time. This intriguing phenomenon, discovered in replay-based Class Incr…