Generalized Leverage Score Sampling for Neural Networks

Authors: Jason D. Lee, Ruoqi Shen, Zhao Song, Mengdi Wang, zheng Yu

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We summarize our main results and contribution as following: Generalize the leverage score sampling theory for kernel ridge regression to a broader class of kernels. Connect the leverage score sampling theory with neural network training. Theoretically prove the equivalence between training regularized neural network and kernel ridge regression under both random Gaussian initialization and leverage score sampling initialization. The focus of this paper is purely theoretical, and thus a broader impact discussion is not applicable.
Researcher Affiliation Academia jasonlee@princeton.edu Princeton University. Work done while visiting Institute for Advanced Study. shenr3@cs.washington.edu University of Washington. Work done while visiting Institute for Advanced Study. zhaos@ias.edu Columbia University, Princeton University and Institute for Advanced Study. mengdiw@princeton.edu Princeton University. zhengy@princeton.edu Princeton University.
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code No The paper states 'The full version of this paper is available at https://arxiv.org/pdf/2009.09829.pdf' which is a link to the paper itself, not to source code. No other statement about code availability is present.
Open Datasets No This paper focuses on theoretical analysis and does not describe experiments involving datasets.
Dataset Splits No This paper is theoretical and does not describe experiments with dataset splits.
Hardware Specification No This is a theoretical paper and does not report on experiments or hardware specifications.
Software Dependencies No This paper is theoretical and does not describe software dependencies or specific versions.
Experiment Setup No This paper is theoretical and does not include details about an experimental setup or hyperparameters.