Learning to Optimize with Dynamic Mode Decomposition

被引:0
|
作者
Simanek, Petr [1 ,2 ]
Vasata, Daniel [1 ]
Kordik, Pavel [1 ]
机构
[1] Czech Tech Univ, Fac Informat Technol, Prague, Czech Republic
[2] GoodAI, Prague, Czech Republic
关键词
machine learning; recurrent neural networks; optimization;
D O I
10.1109/IJCNN55064.2022.9892364
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Designing faster optimization algorithms is of ever-growing interest. In recent years, learning to learn methods that learn how to optimize demonstrated very encouraging results. Current approaches usually do not effectively include the dynamics of the optimization process during training. They either omit it entirely or only implicitly assume the dynamics of an isolated parameter. In this paper, we show how to utilize the dynamic mode decomposition method for extracting informative features about optimization dynamics. By employing those features, we show that our learned optimizer generalizes much better to unseen optimization problems in short. The improved generalization is illustrated on multiple tasks where training the optimizer on one neural network generalizes to different architectures and distinct datasets.
引用
收藏
页数:8
相关论文
共 50 条
  • [1] Deep learning enhanced dynamic mode decomposition
    Alford-Lago, D. J.
    Curtis, C. W.
    Ihler, A. T.
    Issan, O.
    [J]. CHAOS, 2022, 32 (03)
  • [2] Extended Dynamic Mode Decomposition with Invertible Dictionary Learning
    Jin, Yuhong
    Hou, Lei
    Zhong, Shun
    [J]. NEURAL NETWORKS, 2024, 173
  • [3] Randomized Projection Learning Method for Dynamic Mode Decomposition
    Surasinghe, Sudam
    M. Bollt, Erik
    [J]. MATHEMATICS, 2021, 9 (21)
  • [4] Learning Koopman Invariant Subspaces for Dynamic Mode Decomposition
    Takeishi, Naoya
    Kawahara, Yoshinobu
    Yairi, Takehisa
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 30 (NIPS 2017), 2017, 30
  • [5] Supervised dynamic mode decomposition via multitask learning
    Fujii, Keisuke
    Kawahara, Yoshinobu
    [J]. PATTERN RECOGNITION LETTERS, 2019, 122 : 7 - 13
  • [6] Accelerated Reinforcement Learning via Dynamic Mode Decomposition
    Donge, Vrushabh S.
    Lian, Bosen
    Lewis, Frank L.
    Davoudi, Ali
    [J]. IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, 2023, 10 (04): : 2022 - 2034
  • [7] New method for dynamic mode decomposition of flows over moving structures based on machine learning (hybrid dynamic mode decomposition)
    Naderi, Mohammad Hossein
    Eivazi, Hamidreza
    Esfahanian, Vahid
    [J]. PHYSICS OF FLUIDS, 2019, 31 (12)
  • [8] Machine learning enhanced Hankel dynamic-mode decomposition
    Curtis, Christopher W.
    Alford-Lago, D. Jay
    Bollt, Erik
    Tuma, Andrew
    [J]. CHAOS, 2023, 33 (08)
  • [9] A dynamic mode decomposition based deep learning technique for prognostics
    Akkad, Khaled
    He, David
    [J]. JOURNAL OF INTELLIGENT MANUFACTURING, 2023, 34 (05) : 2207 - 2224
  • [10] A dynamic mode decomposition based deep learning technique for prognostics
    Khaled Akkad
    David He
    [J]. Journal of Intelligent Manufacturing, 2023, 34 : 2207 - 2224