On Neural Network Kernels and the Storage Capacity Problem

Author:

Zavatone-Veth Jacob A.1,Pehlevan Cengiz2

Affiliation:

1. Department of Physics and Center for Brain Science, Harvard University, Cambridge, MA 02138, U.S.A. jzavatoneveth@g.harvard.edu

2. Center for Brain Science and John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, U.S.A. cpehlevan@seas.harvard.edu

Abstract

Abstract In this short note, we reify the connection between work on the storage capacity problem in wide two-layer treelike neural networks and the rapidly growing body of literature on kernel limits of wide neural networks. Concretely, we observe that the “effective order parameter” studied in the statistical mechanics literature is exactly equivalent to the infinite-width neural network gaussian process kernel. This correspondence connects the expressivity and trainability of wide two-layer neural networks.

Publisher

MIT Press - Journals

Subject

Cognitive Neuroscience,Arts and Humanities (miscellaneous)

Reference20 articles.

1. Properties of the geometry of solutions and capacity of multilayer neural networks with rectified linear unit activations;Baldassi;Physical Review Letters,2019

2. Broken symmetries in multilayered perceptrons;Barkai;Physical Review A,1992

3. Gaussian Measures

4. Kernel methods for deep learning;Cho,2009

5. Toward deeper understanding of neural networks: The power of initialization and a dual view on expressivity;Daniely,2016

Cited by 1 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

1. Bayesian interpolation with deep linear networks;Proceedings of the National Academy of Sciences;2023-05-30

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3