Concepedia

Gradient Balancing

Gradient Balancing is a multi-task learning strategy that dynamically adjusts the relative influence of task-specific gradients during optimization to mitigate competition among conflicting objectives, thereby improving joint model performance.

19

Publications

1.4K

Citations

95

Authors

34

Institutions

Publications per year

2017–2026

19

Parents

Publications

19

GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks

Zhao Chen, Vijay Badrinarayanan, Chen‐Yu Lee et al. · arXiv (Cornell University) · 2017 · 455 citations · Full text

, , +16

GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks

Chen Zhao, Vijay Badrinarayanan, Chen‐Yu Lee et al. · International Conference on Machine Learning · 2018 · 314 citations

, , +16

Equalization Loss v2: A New Gradient Balance Approach for Long-tailed Object Detection

Jingru Tan, Xin Lu, Gang Zhang et al. · 2021 · 179 citations

, , +14

Dynamic Weight Strategy of Physics-Informed Neural Networks for the 2D Navier–Stokes Equations

Shirong Li, Xinlong Feng · Entropy · 2022 · 70 citations · Full text

, , +17

Just Pick a Sign: Optimizing Deep Multitask Models with Gradient Sign Dropout

Chen Zhao, Jiquan Ngiam, Yanping Huang et al. · arXiv (Cornell University) · 2020 · 64 citations · Full text

, , +15

1–5 of 19

Authors

95

  1. JT
    Jingru Tan

    Tongji University, China

    2 publications · 210 citations

    2

    h-index

  2. YC
    Yuning Chai

    Google (United States), United States

    2 publications · 74 citations

    2

    h-index

  3. TL
    Thang Luong

    Stanford University, United States

    2 publications · 74 citations

    2

    h-index

  4. VB
    Vijay Badrinarayanan

    University of Cambridge, United Kingdom

    2 publications · 769 citations

    2

    h-index

  5. YH
    Yanping Huang

    Hong Kong Polytechnic University, Hong Kong SAR China

    2 publications · 74 citations

    2

    h-index

1–5 of 95

Institutions

34

  1. Google (United States)

    Mountain View, United States

    3 h-index

    9 publications

    556 citations

  2. Tongji University

    Shanghai, China

    2 h-index

    3 publications

    389 citations

  3. Group Sense (China)

    Hong Kong, China

    2 h-index

    5 publications

    145 citations

  4. Tsinghua University

    Beijing, China

    2 h-index

    4 publications

    257 citations

  5. Harbin Institute of Technology

    Harbin, China

    1 h-index

    6 publications

    120 citations

  6. Xinjiang University

    Ürümqi, China

    1 h-index

    2 publications

    140 citations

  7. Shanghai Artificial Intelligence Laboratory

    Shanghai, China

    1 h-index

    2 publications

    62 citations

  8. Jingdong (China)

    Beijing, China

    1 h-index

    1 publications

    53 citations

1–8 of 34

Venues

  1. arXiv (Cornell University)

    Repository · 2 publications · 519 citations

    2

    h-index

  2. UCL Discovery (University College London)

    Repository · 1 publications · 26 citations

    1

    h-index

  3. Machine Vision and Applications

    1 publications · 10 citations

    1

    h-index

  4. Neural Information Processing Systems

    Conference · 1 publications · 10 citations

    1

    h-index

  5. IEEE Transactions on Industrial Informatics

    1 publications · 24 citations

    1

    h-index

1–5