Faster Meta Update Strategy for Noise-Robust Deep Learning

被引:37
|
作者
Xu, Youjiang [1 ]
Zhu, Linchao [2 ]
Jiang, Lu [3 ]
Yang, Yi [2 ]
机构
[1] Baidu Res, Beijing, Peoples R China
[2] Univ Technol Sydney, ReLER, Sydney, NSW, Australia
[3] Google Res, Mountain View, CA USA
关键词
D O I
10.1109/CVPR46437.2021.00021
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
It has been shown that deep neural networks are prone to overfitting on biased training data. Towards addressing this issue, meta-learning employs a meta model for correcting the training bias. Despite the promising performances, super slow training is currently the bottleneck in the meta learning approaches. In this paper, we introduce a novel Faster Meta Update Strategy (FaMUS) to replace the most expensive step in the meta gradient computation with a faster layer-wise approximation. We empirically find that FaMUS yields not only a reasonably accurate but also a low-variance approximation of the meta gradient. We conduct extensive experiments to verify the proposed method on two tasks. We show our method is able to save two-thirds of the training time while still maintaining the comparable or achieving even better generalization performance. In particular, our method achieves the state-of-the-art performance on both synthetic and realistic noisy labels, and obtains promising performance on long-tailed recognition on standard benchmarks.
引用
收藏
页码:144 / 153
页数:10
相关论文
共 50 条
  • [21] Bag-of-Deep-Features: Noise-Robust Deep Feature Representations for Audio Analysis
    Amiriparian, Shahin
    Gerczuk, Maurice
    Ottl, Sandra
    Cummins, Nicholas
    Pugachevskiy, Sergey
    Schuller, Bjoern
    2018 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2018,
  • [22] Switchable K-class Hyperplanes for Noise-Robust Representation Learning
    Liu, Boxiao
    Song, Guanglu
    Zhang, Manyuan
    You, Haihang
    Liu, Yu
    2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021), 2021, : 2999 - 3008
  • [23] Adversarial Multi-Task Deep Learning for Noise-Robust Voice Activity Detection with Low Algorithmic Delay
    Larsen, Claus M.
    Koch, Peter
    Tan, Zheng-Hua
    INTERSPEECH 2022, 2022, : 3759 - 3763
  • [24] A NOISE-ROBUST SIGNAL PROCESSING STRATEGY FOR COCHLEAR IMPLANTS USING NEURAL NETWORKS
    Zheng, Nengheng
    Shi, Yupeng
    Kang, Yuyong
    Meng, Qinglin
    2021 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP 2021), 2021, : 8343 - 8347
  • [25] Noise-Robust Learning from Multiple Unsupervised Sources of Inferred Labels
    Silva, Amila
    Luo, Ling
    Karunasekera, Shanika
    Leckie, Christopher
    THIRTY-SIXTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FOURTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE / TWELVETH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2022, : 8315 - 8323
  • [26] SphereNet: Learning a Noise-Robust and General Descriptor for Point Cloud Registration
    Zhao, Guiyu
    Guo, Zhentao
    Wang, Xin
    Ma, Hongbin
    IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2024, 62 : 1 - 16
  • [27] SYNERGISTIC NETWORK LEARNING AND LABEL CORRECTION FOR NOISE-ROBUST IMAGE CLASSIFICATION
    Gong, Chen
    Bin, Kong
    Seibel, Eric J.
    Wang, Xin
    Yin, Youbing
    Song, Qi
    2022 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 2022, : 4253 - 4257
  • [28] Noise-robust acoustic signature recognition using nonlinear Hebbian learning
    Lu, Bing
    Dibazar, Alireza
    Berger, Theodore W.
    NEURAL NETWORKS, 2010, 23 (10) : 1252 - 1263
  • [29] Noise-robust Attention Learning for End-to-End Speech Recognition
    Higuchi, Yosuke
    Tawara, Naohiro
    Ogawa, Atsunori
    Iwata, Tomoharu
    Kobayashi, Tetsunori
    Ogawa, Tetsuji
    28TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2020), 2021, : 311 - 315
  • [30] Noise-robust deep learning ghost imaging using a non-overlapping pattern for defect position mapping
    Kataoka, Shoma
    Mizutani, Yasuhiro
    Uenohara, Tsutomu
    Takaya, Yasuhiro
    Matoba, Osamu
    APPLIED OPTICS, 2022, 61 (34) : 10126 - 10133