Gradient Accumulation

Gradient Accumulation is a computational technique in deep learning optimization whereby gradients are accumulated over multiple mini-batches before performing a single parameter update, enabling effective training with larger batch sizes under limited memory constraints.

23

Publications

3.6K

Citations

102

Authors

46

Institutions

Publications per year

2017–2026

16

Authors

102

  1. PM
    Paulius Micikevicius

    Nvidia (United Kingdom), United Kingdom

    2 publications · 1.1K citations

    2

    h-index

  2. JL
    Jiadong Lin

    First Affiliated Hospital of Xi'an Jiaotong University, China

    2 publications · 50 citations

    2

    h-index

  3. HW
    Hao Wu

    Xi'an Jiaotong University, China

    2 publications · 1.1K citations

    2

    h-index

  4. EE
    Erich Elsen

    Google (United States), United States

    2 publications · 1.1K citations

    2

    h-index

  5. OK
    Oleksii Kuchaiev

    University of California, Irvine, United States

    2 publications · 1.1K citations

    2

    h-index

1–5 of 102

Venues

  1. arXiv (Cornell University)

    Repository · 5 publications · 1.2K citations

    5

    h-index

  2. 1

    h-index

  3. Cement and Concrete Composites

    1 publications · 30 citations

    1

    h-index

  4. IEEE Transactions on Communications

    1 publications · 81 citations

    1

    h-index

  5. IEEE Internet of Things Journal

    1 publications · 91 citations

    1

    h-index

1–5