arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2002.02561cs.LGstat.ML
收藏

核回归和宽神经网络中的谱依赖学习曲线

Spectrum Dependent Learning Curves in Kernel Regression and Wide Neural Networks

John A. Paulson School of Engineering and Applied Sciences, Harvard University(哈佛大学约翰·A·保尔森工程与应用科学学院);Center for Brain Science, Harvard University(哈佛大学脑科学中心)

Blake Bordelon, Abdulkadir Canatar, Cengiz Pehlevan

摘要:

我们从高斯过程和统计物理的理论方法推导出作为训练样本数量的函数的核回归的泛化性能的解析表达式。我们的表达式适用于广泛的神经网络,由于它们的训练和核回归与神经切线核(NTK)之间的等价性。通过计算由于内核的不同谱分量的总的泛化误差的分解,我们确定了一个新的谱原则:随着训练集的大小的增长,内核机器和神经网络适合连续更高的谱模式的目标函数。当数据从高维超球面上的均匀分布中采样时,包括NTK在内的点积核表现出学习阶段,其中学习目标函数的不同频率模式。我们用合成数据和MNIST数据集验证了我们的理论。

英文摘要:

We derive analytical expressions for the generalization performance of kernel regression as a function of the number of training samples using theoretical methods from Gaussian processes and statistical physics. Our expressions apply to wide neural networks due to an equivalence between training them and kernel regression with the Neural Tangent Kernel (NTK). By computing the decomposition of the total generalization error due to different spectral components of the kernel, we identify a new spectral principle: as the size of the training set grows, kernel machines and neural networks fit successively higher spectral modes of the target function. When data are sampled from a uniform distribution on a high-dimensional hypersphere, dot product kernels, including NTK, exhibit learning stages where different frequency modes of the target function are learned. We verify our theory with simulations on synthetic data and MNIST dataset.

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。

↑