Self-Consistent Learning of Neural Dynamical Systems From Noisy Time Series

被引:9
作者
Wang, Zhe [1 ]
Guet, Claude [2 ,3 ]
机构
[1] CNRS CREATE, Singapore 138602, Singapore
[2] Nanyang Technol Univ, Energy Res Inst NTU, Singapore 637141, Singapore
[3] Nanyang Technol Univ, Sch Phys & Math Sci, Singapore 637371, Singapore
来源
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE | 2022年 / 6卷 / 05期
基金
新加坡国家研究基金会;
关键词
Time series analysis; Dynamical systems; Neural networks; Mathematical models; Delays; Time measurement; Noise measurement; Denoising and deterministic forecasting; neural dynamical systems; self-consistent learning; PREDICTION; DELAY; NETWORK;
D O I
10.1109/TETCI.2022.3146332
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We introduce a new method which, for a single noisy time series, provides unsupervised filtering, state space reconstruction, efficient learning of the unknown governing multivariate dynamical system, and deterministic forecasting. We construct both the underlying trajectories and a latent dynamical system using deep neural networks. Under the assumption that the trajectories follow the latent dynamical system, we determine the unknowns of the dynamical system, and filter out stochastic outliers in the measurements. In this sense the method is self-consistent. The embedding dimension is determined iteratively during the training by using the false-nearest-neighbors Algorithm and it is implemented as an attention map to the state vector. This allows for a state space reconstruction without a priori information on the signal. By exploiting the differentiability of the neural solution trajectory, we can define the neural dynamical system locally at each time, mitigating the need for forward and backwards passing through numerical solvers of the canonical adjoint method. On a chaotic time series masked by additive Gaussian noise, we demonstrate that the denoising ability and the predictive power of the proposed method are mainly due to the self-consistency, insensitive to methods used for the state space reconstruction.
引用
收藏
页码:1103 / 1112
页数:10
相关论文
共 50 条
[41]   Reconstructing bifurcation diagrams of dynamical systems using measured time series [J].
Bagarinao, E ;
Pakdaman, K ;
Nomura, T ;
Sato, S .
METHODS OF INFORMATION IN MEDICINE, 2000, 39 (02) :146-149
[42]   Hierarchical Linear Dynamical Systems: A new model for clustering of time series [J].
Cinar, Goktug T. ;
Loza, Carlos A. ;
Principe, Jose C. .
PROCEEDINGS OF THE 2014 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2014, :2464-2470
[43]   Modeling Spatio-Temporal Dynamical Systems With Neural Discrete Learning and Levels-of-Experts [J].
Wang, Kun ;
Wu, Hao ;
Zhang, Guibin ;
Fang, Junfeng ;
Liang, Yuxuan ;
Wu, Yuankai ;
Zimmermann, Roger ;
Wang, Yang .
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2024, 36 (08) :4050-4062
[44]   Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects [J].
Zhang, Kexin ;
Wen, Qingsong ;
Zhang, Chaoli ;
Cai, Rongyao ;
Jin, Ming ;
Liu, Yong ;
Zhang, James Y. ;
Liang, Yuxuan ;
Pang, Guansong ;
Song, Dongjin ;
Pan, Shirui .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2024, 46 (10) :6775-6794
[45]   Forecasting Financial Time Series Using a Hybrid Self-Organising Neural Model [J].
Ouyang, Yicun ;
Yin, Hujun .
INTELLIGENT DATA ENGINEERING AND AUTOMATED LEARNING - IDEAL 2013, 2013, 8206 :262-269
[46]   Learning Dynamical Systems by Leveraging Data From Similar Systems [J].
Xin, Lei ;
Ye, Lintao ;
Chiu, George ;
Sundaram, Shreyas .
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2025, 70 (07) :4833-4840
[47]   Dynamic Learning From Adaptive Neural Control for Discrete-Time Strict-Feedback Systems [J].
Wang, Min ;
Shi, Haotian ;
Wang, Cong ;
Fu, Jun .
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2022, 33 (08) :3700-3712
[48]   Prediction of noisy chaotic time series using an optimal radial basis function neural network [J].
Leung, H ;
Lo, T ;
Wang, SC .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 2001, 12 (05) :1163-1172
[49]   SCAN-SPECIFIC ACCELERATED MRI RECONSTRUCTION USING RECURRENT NEURAL NETWORKS IN A REGULARIZED SELF-CONSISTENT FRAMEWORK [J].
Hosseini, Seyed Amir Hossein ;
Yaman, Burhaneddin ;
Zhang, Chi ;
Ukurbil, Kamil ;
Moeller, Steen ;
Akcakaya, Mehmet .
2020 IEEE 17TH INTERNATIONAL SYMPOSIUM ON BIOMEDICAL IMAGING WORKSHOPS (IEEE ISBI WORKSHOPS 2020), 2020,
[50]   RESEARCH ON PARAMETER IDENTIFICATION OF COMPOSITE MATERIALS BY COMBINATION OF SELF-CONSISTENT CLUSTER ANALYSIS, NEURAL NETWORK AND BAYESIAN OPTIMIZATION [J].
Feng, Yixin ;
Peng, Hui ;
Luo, Wei .
Lixue Xuebao/Chinese Journal of Theoretical and Applied Mechanics, 2024, 56 (11) :3333-3350