Concepedia

Movement Pruning

Movement Pruning is a structured sparsity method in deep neural networks that identifies and removes weights with the smallest "movement" values—computed as the product of weight magnitude and gradient sign—during training to achieve efficient model compression while preserving performance.

1

Publications

33

Citations

3

Authors

2

Institutions

Publications per year

2017–2026

1

Parents

Publications

1

1–2 of 1

Authors

3

  1. VS
    Victor Sanh

    Hugging Face, United States

    1 publications · 33 citations

    1

    h-index

  2. AM
    Alexander M. Rush

    Harvard University Press, United States

    1 publications · 33 citations

    1

    h-index

  3. TW
    Thomas Wolf

    Louisiana State University, United States

    1 publications · 33 citations

    1

    h-index

  4. AA
    Akiko Aizawa

    National Institute of Informatics, Japan

    1 publications · 12 citations

    0

    h-index

  5. PJ
    Przemyslaw Joniak

    The University of Tokyo, Japan

    1 publications · 12 citations

    0

    h-index

1–5 of 3

Institutions

2

  1. Central European University

    Vienna, Austria

    1 h-index

    1 publications

    33 citations

  2. Cornell University

    Ithaca, United States

    1 h-index

    1 publications

    33 citations

  3. The University of Tokyo

    Tokyo, Japan

    1 publications

    12 citations

  4. National Institute of Informatics

    Tokyo, Japan

    1 publications

    12 citations

1–4 of 2

Venues

  1. arXiv (Cornell University)

    Repository · 1 publications · 33 citations

    1

    h-index

1–1