提出轻量级网格采样法，高效压缩记忆样本提升在线增量学习性能
GPS: Distilling Compact Memories via Grid-based Patch Sampling for Efficient Online Class-Incremental Learning
- 通过网格化像素采样生成低分辨率紧凑表示，无需训练模型
- 在内存受限下实现3%-4%的平均准确率提升，计算开销小
- 适合资源受限场景下的持续学习系统部署
在线类增量学习旨在让模型在仅有限访问历史数据的情况下持续适应新类别，同时缓解灾难性遗忘。基于重放的方法通过维护一个小型历史样本缓冲区来应对这一挑战，表现优异。为在存储受限条件下提高重放效率，近期方法采用知识蒸馏生成更具信息量的记忆样本，但常伴随显著的计算开销，因依赖双层优化。为此，我们提出网格化补丁采样（GPS），一种无需可训练模型的轻量级、高效的记忆样本蒸馏策略。GPS通过从原图中采样部分像素，生成保留语义与结构信息的紧凑低分辨率表示。重放时，这些表示被重建以支持训练与评估。在多个基准上的实验表明，GPS可无缝集成至现有重放框架，在内存受限条件下实现3%-4%的平均端到端准确率提升，且计算开销极低。
原文摘要 · Abstract (English)
Online class-incremental learning aims to enable models to continuously adapt to new classes with limited access to past data, while mitigating catastrophic forgetting. Replay-based methods address this by maintaining a small memory buffer of previous samples, achieving competitive performance. For effective replay under constrained storage, recent approaches leverage distilled data to enhance the informativeness of memory. However, such approaches often involve significant computational overhead due to the use of bi-level optimization. Motivated by these limitations, we introduce Grid-based Patch Sampling (GPS), a lightweight and effective strategy for distilling informative memory samples without relying on a trainable model. GPS generates informative samples by sampling a subset of pixels from the original image, yielding compact low-resolution representations that preserve both semantic content and structural information. During replay, these representations are reassembled to support training and evaluation. Experiments on extensive benchmarks demonstrate that GRS can be seamlessly integrated into existing replay frameworks, leading to 3%-4% improvements in average end accuracy under memory-constrained settings, with limited computational overhead.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。