📂读读paper

Shi, Guangyuan, et al. "Overcoming catastrophic forgetting in incremental few-shot learning by finding flat minima." Advances in neural information processing systems 34 (2021): 6747-6761.code available论文关注Continual Learning中的Few-shot问题，发现对于few-shot问题，一个简单的基类训练的基线模型优于最先进的方法，这证明了灾难性遗忘的严重性。论文提出了一种寻找到基

Islam, Ashraful, et al. "A broad study on the transferability of visual representations with contrastive learning." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021.code available论文主要目标是了解对比学习优越于标准监督交叉熵模型的可迁移性的潜在机制。Contributions*1.Benchmark five methods, whose results sh

Kornblith, Simon, et al. "Similarity of neural network representations revisited." International conference on machine learning. PMLR, 2019.No open source. this is the reproduce of others. Seeking from other papers might be accessible.论文主要引入了一种CKA的度量方式，用以确定从不同的随机初始化和不同的宽度下训练出来的神经网络的隐藏层之间的对应关系。验证了更宽的

Madaan, Divyam, et al. "Representational continuity for unsupervised continual learning." arXiv preprint arXiv:2110.06976 (2021). ICLR2022code available论文通过严谨的实验观察到无监督表征比监督表征对灾难性遗忘更加健壮。微调任务的顺序可以使得效果超过sota的持续学习模型。通过CKA分析可以得到UCL与SCL在低级层的学习表征大致相同，但是在高级层的学习表征有很大不同，这可能导致了对灾难性遗忘抵御效果的差异。Contributions*1.at

Kim, Sanghwan, et al. "Achieving a Better Stability-Plasticity Trade-off via Auxiliary Networks in Continual Learning." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023.code available_论文中提到有多种方式定义持续学习的问题，分为3类，任务增量学习（TIL）、领域增量学习（DIL）和类增量学习（CIL）。在TIL中，模型被告知需要解决的任