让模型自己筛选干净数据，训练更稳更准。
Step-E: A Differentiable Data Cleaning Framework for Robust Learning with Noisy Labels
- 模型每轮自动排除高损失样本，动态优化学习过程。
- 在CIFAR-100N上将准确率从43.3%提升至50.4%，接近60.5%的纯净数据上限。
- 无需预处理，适合真实场景中标签混乱的数据集使用。
野外采集的训练数据常含噪声标签和异常值，严重损害深度神经网络的性能与可靠性。传统数据清洗作为独立预处理阶段，无法利用下游模型反馈，也难以适应未知噪声模式。本文提出Step-E框架，将样本选择与模型学习整合为单一优化过程。每轮训练中，Step-E根据损失对样本排序，并在短暂预热后逐步剔除高损失样本的梯度更新，形成在线课程，聚焦于易学且一致的样本，最终忽略顽固异常值。在CIFAR-100N上，Step-E将ResNet-18测试准确率从43.3%（±0.7%）提升至50.4%（±0.9%），显著优于损失截断、自定步长学习和一次性过滤方法，接近60.5%（±0.2%）的纯净标签理想上限。在CIFAR-10N（aggre）上，也从83.9%提升至85.3%，接近85.9%的纯净标签基准，仅需适度训练开销。
原文摘要 · Abstract (English)
Training data collected in the wild often contain noisy labels and outliers that substantially degrade the performance and reliability of deep neural networks. While data cleaning is commonly applied as a separate preprocessing stage, such two-stage pipelines neither fully exploit feedback from the downstream model nor adapt to unknown noise patterns. We propose Step-E, a simple framework that integrates sample selection and model learning into a single optimization process. At each epoch, Step-E ranks samples by loss and gradually increases the fraction of high-loss examples that are excluded from gradient updates after a brief warm-up stage, yielding an online curriculum that focuses on easy and consistent examples and eventually ignores persistent outliers. On CIFAR-100N, Step-E improves the test accuracy of a ResNet-18 model from 43.3% (+/- 0.7%) to 50.4% (+/- 0.9%), clearly outperforming loss truncation, self-paced learning, and one-shot filtering while approaching the clean-label oracle at 60.5% (+/- 0.2%). On CIFAR-10N (aggre), Step-E also improves over the noisy baseline (85.3% vs. 83.9%) and nearly matches the clean-label oracle (85.9%), with only moderate training-time 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。