Spectral bias and task-model alignment explain generalization in kernel regression and infinitely wide neural networks.
Spectral bias and task-model alignment explain generalization in kernel regression and infinitely wide neural networks.
复制标题
DOI:
10.1038/s41467-021-23103-1
复制
发表时间:
2021-05-18
影响因子:
16.6
通讯作者:
Pehlevan C
中科院分区:
文献类型:
--
作者:
Canatar A;Bordelon B;Pehlevan C
A theoretical understanding of generalization remains an open problem for many machine learning models, including deep networks where overparameterization leads to better performance, contradicting the conventional wisdom from classical statistics. Here, we investigate generalization error for kernel regression, which, besides being a popular machine learning method, also describes certain infinitely overparameterized neural networks. We use techniques from statistical mechanics to derive an analytical expression for generalization error applicable to any kernel and data distribution. We present applications of our theory to real and synthetic datasets, and for many kernels including those that arise from training deep networks in the infinite-width limit. We elucidate an inductive bias of kernel regression to explain data with simple functions, characterize whether a kernel is compatible with a learning task, and show that more data may impair generalization when noisy or not expressible by the kernel, leading to non-monotonic learning curves with possibly many peaks. Canatar et al. propose a predictive theory of generalization in kernel regression applicable to real data. This theory explains various generalization phenomena observed in wide neural networks, which admit a kernel limit and generalize well despite being overparameterized.
登录
查看更多内容
影响因子:
3
作者:
Cucker, F;Smale, S
通讯作者:
Smale, S
影响因子:
12.5
作者:
Advani, Madhu;Ganguli, Surya
通讯作者:
Ganguli, Surya
DOI:
10.1088/1742-5468/2005/05/p05012
发表时间:
2005-05-01
影响因子:
2.4
作者:
Castellani, T;Cavagna, A
通讯作者:
Cavagna, A
DOI:
10.1073/pnas.1903070116
发表时间:
2019-08-06
影响因子:
11.1
作者:
Belkin, Mikhail;Hsu, Daniel;Mandal, Soumik
通讯作者:
Mandal, Soumik
影响因子:
8.6
作者:
Dietrich, R;Opper, M;Sompolinsky, H
通讯作者:
Sompolinsky, H