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 2102.00200cs.LGphysics.data-an
收藏

线性频率原理模型解释神经网络中的过拟合缺失

Linear Frequency Principle Model to Understand the Absence of Overfitting in Neural Networks

Shanghai Jiao Tong University(上海交通大学)

Yaoyu Zhang, Tao Luo, Zheng Ma, Zhi-Qin John Xu

摘要:

为什么高度参数化的神经网络(NN)不会过拟合数据是一个重要的长期悬而未决的问题。我们提出了一个现象学模型的NN训练来解释这个非过拟合的难题。我们的线性频率原理(LFP)模型解释了NN的一个关键动力学特征:它们首先学习低频,而不管微观细节。基于LFP模型的理论表明,目标函数的低频优势是神经网络不过拟合的关键条件,并通过实验得到了验证。此外,通过一个理想的两层NN,我们揭示了详细的微观NN训练动态如何在统计上产生具有定量预测能力的LFP模型。

英文摘要:

Why heavily parameterized neural networks (NNs) do not overfit the data is an important long standing open question. We propose a phenomenological model of the NN training to explain this non-overfitting puzzle. Our linear frequency principle (LFP) model accounts for a key dynamical feature of NNs: they learn low frequencies first, irrespective of microscopic details. Theory based on our LFP model shows that low frequency dominance of target functions is the key condition for the non-overfitting of NNs and is verified by experiments. Furthermore, through an ideal two-layer NN, we unravel how detailed microscopic NN training dynamics statistically gives rise to a LFP model with quantitative prediction power.

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。

↑