arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2302.01952stat.MLcs.LG
收藏

论深度学习中梯度下降动力学与不稳定性的连续时间模型

On a continuous time model of gradient descent dynamics and instability in deep learning

DeepMind(深度思维公司);University College London(伦敦大学学院);Google(谷歌公司)

Mihaela Rosca, Yan Wu, Chongli Qin, Benoit Dherin

AI总结:

为研究深度学习中梯度下降的不稳定性,本文提出连续时间流 PF 近似其动力学,揭示边缘稳定性现象,并据此提出一种学习率自适应方法以权衡训练稳定性与测试性能。

摘要:

深度学习成功的秘诀在于神经网络与基于梯度的优化的结合。然而,对梯度下降行为(尤其是其不稳定性)的理解,却落后于其经验上的成功。为了丰富研究梯度下降的理论工具,我们提出了 principal flow (PF),一种用于近似梯度下降动力学的连续时间流。据我们所知,PF 是唯一能够捕捉梯度下降发散与振荡行为的连续流,包括逃离局部极小值和鞍点。通过依赖于 Hessian 矩阵的特征分解,PF 揭示了深度学习中近期观察到的边缘稳定性现象。基于对不稳定性的全新理解,我们提出了一种学习率自适应方法,使我们能够控制训练稳定性与测试集评估性能之间的权衡。

英文摘要:

The recipe behind the success of deep learning has been the combination of neural networks and gradient-based optimization. Understanding the behavior of gradient descent however, and particularly its instability, has lagged behind its empirical success. To add to the theoretical tools available to study gradient descent we propose the principal flow (PF), a continuous time flow that approximates gradient descent dynamics. To our knowledge, the PF is the only continuous flow that captures the divergent and oscillatory behaviors of gradient descent, including escaping local minima and saddle points. Through its dependence on the eigendecomposition of the Hessian the PF sheds light on the recently observed edge of stability phenomena in deep learning. Using our new understanding of instability we propose a learning rate adaptation method which enables us to control the trade-off between training stability and test set evaluation performance.

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。

↑