Matteo Boschini, Lorenzo Bonicelli, Angelo Porrello, Giovanni Bellitto, Matteo Pennisi...
TL;DR本文研究了连续学习和迁移学习之间的纠缠关系，提出了一种新方法Transfer without Forgetting来解决预训练网络在后续任务中遗忘知识的问题，并证明其在各种数据集和不同缓冲区大小下都能显着优于其他连续学习方法，平均提高了4.81％的类增量准确性。
Abstract
This work investigates the entanglement between continual learning (CL) and transfer learning (TL). In particular, we shed light on the widespread application of →