arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.13646cs.CVcs.LG
收藏

对比Divide:带噪声标签学习的自我监督预训练

Contrast to Divide: Self-Supervised Pre-Training for Learning with Noisy Labels

Technion -- Israel Institute of Technology(以色列理工学院);NVIDIA(英伟达)

Evgenii Zheltonozhskii, Chaim Baskin, Avi Mendelson, Alex M. Bronstein, Or Litany

摘要:

使用噪声标签(LNL)方法学习的成功在很大程度上依赖于预热阶段的成功,其中使用完整(噪声)训练集执行标准监督训练。在本文中,我们确定了一个“热身障碍”:标准的热身阶段无法训练高质量的特征提取器,避免记忆的嘈杂标签。我们提出了“对比划分”(C2 D),这是一个简单的框架,通过以自我监督的方式预训练特征提取器来解决这个问题。使用自我监督的预训练通过大幅降低预热阶段对噪声水平的敏感性,缩短其持续时间并提高提取的特征质量来提高现有LNL方法的性能。C2 D与现有方法一起开箱即用,并表现出显着的性能改进,特别是在高噪声区域,我们在CIFAR-100中获得了超过27%的提升,噪声为90%,超过了以前的最先进水平。在现实生活中的噪声设置中,在mini-WebVision上训练的C2 D在WebVision和ImageNet验证集上的表现都优于以前的作品3%的top-1准确度。我们进行了深入的分析框架,包括调查不同的预训练方法的性能和估计的有效上限的LNL性能与半监督学习。复制我们实验的代码可在https://github.com/ContrastToDivide/C2D上获得

英文摘要:

The success of learning with noisy labels (LNL) methods relies heavily on the success of a warm-up stage where standard supervised training is performed using the full (noisy) training set. In this paper, we identify a "warm-up obstacle": the inability of standard warm-up stages to train high quality feature extractors and avert memorization of noisy labels. We propose "Contrast to Divide" (C2D), a simple framework that solves this problem by pre-training the feature extractor in a self-supervised fashion. Using self-supervised pre-training boosts the performance of existing LNL approaches by drastically reducing the warm-up stage's susceptibility to noise level, shortening its duration, and improving extracted feature quality. C2D works out of the box with existing methods and demonstrates markedly improved performance, especially in the high noise regime, where we get a boost of more than 27% for CIFAR-100 with 90% noise over the previous state of the art. In real-life noise settings, C2D trained on mini-WebVision outperforms previous works both in WebVision and ImageNet validation sets by 3% top-1 accuracy. We perform an in-depth analysis of the framework, including investigating the performance of different pre-training approaches and estimating the effective upper bound of the LNL performance with semi-supervised learning. Code for reproducing our experiments is available at https://github.com/ContrastToDivide/C2D

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。

↑