NAPA-VQ：用于持续学习的邻域感知原型增强与向量量化
NAPA-VQ: Neighborhood Aware Prototype Augmentation with Vector Quantization for Continual Learning
AI总结：
针对非样本类别增量学习中的新旧类别特征重叠问题，提出NAPA-VQ框架，利用Neural Gas建模邻域关系并增强旧类原型，在三个基准上显著提升准确率并降低遗忘。
摘要：
灾难性遗忘，即在获取新知识时丧失旧知识，是深度神经网络在现实应用中面临的一个陷阱。许多解决该问题的主流方案依赖于存储样本（先前遇到的数据），但在存在内存限制或隐私约束的应用中，这可能并不可行。因此，近期的关注点集中在基于非样本的类别增量学习（NECIL）上，其中模型在不使用任何过去样本的情况下增量学习新类别。然而，由于缺少旧数据，NECIL方法难以区分旧类别和新类别，导致其特征表示发生重叠。我们提出了NAPA-VQ：带向量量化的邻域感知原型增强，这是一个减少NECIL中此类类别重叠的框架。我们从Neural Gas中汲取灵感，学习特征空间中的拓扑关系，识别最有可能相互混淆的相邻类别。该邻域信息被用于强化相邻类别之间的强分离，并生成旧类别的代表性原型，从而更好地帮助在旧类别与新类别之间获得具有判别性的决策边界。我们在CIFAR-100、TinyImageNet和ImageNet-Subset上的综合实验表明，NAPA-VQ优于最先进的NECIL方法，准确率分别平均提升5%、2%和4%，遗忘程度分别平均降低10%、3%和9%。我们的代码可在https://github.com/TamashaM/NAPA-VQ.git获取。
英文摘要：
Catastrophic forgetting; the loss of old knowledge upon acquiring new knowledge, is a pitfall faced by deep neural networks in real-world applications. Many prevailing solutions to this problem rely on storing exemplars (previously encountered data), which may not be feasible in applications with memory limitations or privacy constraints. Therefore, the recent focus has been on Non-Exemplar based Class Incremental Learning (NECIL) where a model incrementally learns about new classes without using any past exemplars. However, due to the lack of old data, NECIL methods struggle to discriminate between old and new classes causing their feature representations to overlap. We propose NAPA-VQ: Neighborhood Aware Prototype Augmentation with Vector Quantization, a framework that reduces this class overlap in NECIL. We draw inspiration from Neural Gas to learn the topological relationships in the feature space, identifying the neighboring classes that are most likely to get confused with each other. This neighborhood information is utilized to enforce strong separation between the neighboring classes as well as to generate old class representative prototypes that can better aid in obtaining a discriminative decision boundary between old and new classes. Our comprehensive experiments on CIFAR-100, TinyImageNet, and ImageNet-Subset demonstrate that NAPA-VQ outperforms the State-of-the-art NECIL methods by an average improvement of 5%, 2%, and 4% in accuracy and 10%, 3%, and 9% in forgetting respectively. Our code can be found in https://github.com/TamashaM/NAPA-VQ.git.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
