Concepedia

Overparameterization

Overparameterization is a modeling condition in which a statistical or machine learning model possesses significantly more parameters than necessary to fit the underlying data structure, thereby increasing its capacity to memorize noise and potentially compromising generalization despite achieving low training error.

139

Publications

8K

Citations

328

Authors

142

Institutions

Publications per year

2017–2026

112

Parents

Publications

139

A Simple, Fast, and Effective Reparameterization of IBM Model 2

Chris Dyer, Victor Chahuneau, Noah A. Smith · Research Showcase @ Carnegie Mellon University (Carnegie Mellon University) · 2018 · 839 citations

, , +13

Learning Overparameterized Neural Networks via Stochastic Gradient Descent on Structured Data

Yuanzhi Li, Yingyu Liang · arXiv (Cornell University) · 2018 · 319 citations · Full text

, , +16

Gradient descent optimizes over-parameterized deep ReLU networks

Difan Zou, Yuan Cao, Dongruo Zhou et al. · Machine Learning · 2019 · 226 citations · Full text

, , +13

Towards Understanding the Role of Over-Parametrization in Generalization of Neural Networks

Behnam Neyshabur, Zhiyuan Li, Srinadh Bhojanapalli et al. · arXiv (Cornell University) · 2018 · 215 citations · Full text

, , +14

Learning and Generalization in Overparameterized Neural Networks, Going Beyond Two Layers

Zeyuan Allen-Zhu · arXiv (Cornell University) · 2018 · 172 citations · Full text

, , +16

1–5 of 139

Authors

328

  1. MS
    Mahdi Soltanolkotabi

    University of Southern California, United States

    8 publications · 585 citations

    8

    h-index

  2. SO
    Samet Oymak

    California Institute of Technology, United States

    7 publications · 567 citations

    7

    h-index

  3. MB
    Mikhail Belkin

    The Ohio State University, United States

    6 publications · 267 citations

    6

    h-index

  4. QG
    Quanquan Gu

    University of California, Los Angeles, United States

    5 publications · 398 citations

    5

    h-index

  5. FD
    Feng Ding

    Jiangnan University, China

    4 publications · 201 citations

    4

    h-index

1–5 of 328

Institutions

142

  1. Google (United States)

    Mountain View, United States

    11 h-index

    18 publications

    753 citations

  2. University of California, Berkeley

    Berkeley, United States

    11 h-index

    19 publications

    1.2K citations

  3. Stanford University

    Stanford, United States

    6 h-index

    13 publications

    1.2K citations

  4. University of Southern California

    Los Angeles, United States

    6 h-index

    9 publications

    376 citations

  5. Princeton University

    Princeton, United States

    5 h-index

    5 publications

    369 citations

  6. University of California, Los Angeles

    Los Angeles, United States

    5 h-index

    12 publications

    1.2K citations

  7. Jiangnan University

    Wuxi, China

    4 h-index

    10 publications

    557 citations

  8. University of Wisconsin–Madison

    Madison, United States

    4 h-index

    8 publications

    684 citations

1–8 of 142

Venues

  1. arXiv (Cornell University)

    Repository · 51 publications · 2.8K citations

    29

    h-index

  2. Neural Information Processing Systems

    Conference · 4 publications · 252 citations

    4

    h-index

  3. International Journal of Robust and Nonlinear Control

    IJRCEA · 4 publications · 201 citations

    4

    h-index

  4. SIAM Journal on Mathematics of Data Science

    SIMODS · 3 publications · 50 citations

    3

    h-index

  5. 3

    h-index

1–5