Dynamic Sparse Training

Dynamic Sparse Training is a machine learning methodology that iteratively prunes and repurposes network weights during training to maintain sparsity without compromising model performance, distinguished by its adaptive reallocation of connectivity in contrast to static sparse training approaches.

30

Publications

1.4K

Citations

118

Authors

39

Institutions

Publications per year

2017–2026

28

Authors

118

  1. DC
    Decebal Constantin Mocanu

    Eindhoven University of Technology, Netherlands

    6 publications · 123 citations

    6

    h-index

  2. MP
    Mykola Pechenizkiy

    Eindhoven University of Technology, Netherlands

    6 publications · 123 citations

    6

    h-index

  3. EM
    Elena Mocanu

    Eindhoven University of Technology, Netherlands

    3 publications · 50 citations

    3

    h-index

  4. SL
    Shiwei Liu

    Huazhong University of Science and Technology, China

    2 publications · 25 citations

    2

    h-index

  5. SU
    Sebastian U. Stich

    École Polytechnique Fédérale de Lausanne, Switzerland

    2 publications · 102 citations

    2

    h-index

1–5 of 118

Venues

  1. arXiv (Cornell University)

    Repository · 9 publications · 395 citations

    9

    h-index

  2. Proceedings of the AAAI Conference on Artificial Intelligence

    Conference · 2 publications · 129 citations

    2

    h-index

  3. Applied Soft Computing

    1 publications · 13 citations

    1

    h-index

  4. 1

    h-index

  5. 1

    h-index

1–5