arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2406.00396cs.LGcond-mat.stat-mechcs.AIstat.ML
收藏

随机重置缓解标签噪声导致的 SGD 潜在梯度偏差

Stochastic Resetting Mitigates Latent Gradient Bias of SGD from Label Noise

Seoul National University(首尔大学);KAIST(韩国科学技术院)

Youngkyoung Bae, Yeongwoo Song, Hawoong Jeong

AI总结:

本文针对噪声标签使 SGD 产生潜在梯度偏差并损害泛化的问题,引入统计物理中的随机重置机制,理论界定其有效条件并实验验证可显著提升泛化且易于兼容现有方法。

摘要:

在搜索目标或训练深度神经网络(DNN)等许多情形中,放弃并重新开始似乎是一种浪费。然而,我们的研究表明,在使用带噪声标签训练 DNN 时,从检查点重置可以显著提升泛化性能。在存在噪声标签的情况下,DNN 最初会学习数据的一般模式,但随后会逐渐记忆被污染的数据,从而导致过拟合。通过解构随机梯度下降(SGD)的动力学,我们识别出由噪声标签诱发的潜在梯度偏差行为,这种偏差会损害泛化。为缓解这一负面影响,我们将随机重置方法应用于 SGD,其灵感来自统计物理学领域近期在实现高效目标搜索方面的进展。我们首先在理论上确定重置变得有益的条件,随后通过实验验证我们的理论,确认重置带来了显著改进。我们进一步证明,该方法既易于实现,又能与其他处理噪声标签的方法兼容。此外,这项工作还从可解释性视角为理解 DNN 的学习动力学提供了洞见,拓展了借助统计物理学视角分析训练方法的可能性。

英文摘要:

Giving up and starting over may seem wasteful in many situations such as searching for a target or training deep neural networks (DNNs). Our study, though, demonstrates that resetting from a checkpoint can significantly improve generalization performance when training DNNs with noisy labels. In the presence of noisy labels, DNNs initially learn the general patterns of the data but then gradually memorize the corrupted data, leading to overfitting. By deconstructing the dynamics of stochastic gradient descent (SGD), we identify the behavior of a latent gradient bias induced by noisy labels, which harms generalization. To mitigate this negative effect, we apply the stochastic resetting method to SGD, inspired by recent developments in the field of statistical physics achieving efficient target searches. We first theoretically identify the conditions where resetting becomes beneficial, and then we empirically validate our theory, confirming the significant improvements achieved by resetting. We further demonstrate that our method is both easy to implement and compatible with other methods for handling noisy labels. Additionally, this work offers insights into the learning dynamics of DNNs from an interpretability perspective, expanding the potential to analyze training methods through the lens of statistical physics.

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。

↑