Meta-Learning of Neural State-Space Models Using Data From Similar Systems
Not on the map. Why?
Abstract (arXiv)
Deep neural state-space models (SSMs) provide a powerful tool for modeling dynamical systems solely using operational data. Typically, neural SSMs are trained using data collected from the actual system under consideration, despite the likely existence of operational data from similar systems which have previously been deployed in the field. In this paper, we propose the use of model-agnostic meta-learning (MAML) for constructing deep encoder network-based SSMs, by leveraging a combination of archived data from similar systems (used to meta-train offline) and limited data from the actual system (used for rapid online adaptation). We demonstrate using a numerical example that meta-learning can result in more accurate neural SSM models than supervised- or transfer-learning, despite few adaptation steps and limited online data. Additionally, we show that by carefully partitioning and adapting the encoder layers while fixing the state-transition operator, we can achieve comparable performance to MAML while reducing online adaptation complexity.
Key facts. Cited by 12 papers in the map (data through 14 Oct 2026). Builds on 18 of its 25 references that are in the map. First posted on arXiv 14 Nov 2022.
Builds on
18 of 25 references.
References · 18
- 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
- 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,897
- 2017Deep learning for universal linear embeddings of nonlinear dynamicscited by 494
- 2019Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAMLcited by 359
- 2018How to train your MAMLcited by 309
- 2020learn2learn: A Library for Meta-Learning Researchcited by 69
- 2021Adaptive-Control-Oriented Meta-Learning for Nonlinear Systemscited by 39
- 2021Constructing Neural Network-Based Models for Simulating Dynamical Systemscited by 31
- 2020Nonlinear state-space identification using deep encoder networkscited by 30
- 2020dynoNet: a neural network architecture for learning dynamical systemscited by 29
- 2019Model structures and fitting criteria for system identification with neural networkscited by 16
- 2022Identifying the Dynamics of a System by Leveraging Data from Similar Systemscited by 15
6 more
- 2021Deep Identification of Nonlinear Systems in Koopman Formcited by 11
- 2020Physics-Informed Neural State Space Models via Learning and Evolutioncited by 9
- 2020Meta Learning MPC using Finite-Dimensional Gaussian Process Approximationscited by 8
- 2022Learning neural state-space models: do we need a state estimator?cited by 6
- 2022Optimizing Closed-Loop Performance with Data from Similar Systems: A Bayesian Meta-Learning Approachcited by 5
- 2020Meta-Learning Guarantees for Online Receding Horizon Learning Controlcited by 2
Built on by
Citing papers per year: 2023: 2; 2024: 3; 2025: 2; 2026: 5.
12 papers.
- 2023From system models to class models: An in-context learning paradigmcited by 16
- 2023On the adaptation of in-context learners for system identificationcited by 7
- 2024Enhanced Transformer architecture for in-context learning of dynamical systemscited by 7
- 2024Learning Neural Differential Algebraic Equations via Operator Splittingcited by 6
- 2024MPC of Uncertain Nonlinear Systems with Meta-Learning for Fast Adaptation of Neural Predictive Modelscited by 4
- 2025Distributionally robust minimization in meta-learning for system identificationcited by 3
- 2025Manifold meta-learning for reduced-complexity neural system identificationcited by 2
- 2026Physics-Aligned Spectral Mamba: Decoupling Semantics and Dynamics for Few-Shot Hyperspectral Target Detectioncited by 0
- 2026Meta-Learning for Rapid Adaptation in Reference Tracking of Uncertain Nonlinear Systemscited by 0
- 2026T2S-MPC: Time-Embedded Online Adaptive Model Predictive Control for Time-Varying Dynamicscited by 0
- 2026Online learning of neural state-space modelscited by 0
- 2026In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validationcited by 0
- 2026In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validation9 Oct 2026
- 2026Online learning of neural state-space models20 Jul 2026
- 2026T2S-MPC: Time-Embedded Online Adaptive Model Predictive Control for Time-Varying Dynamics24 May 2026
- 2026Meta-Learning for Rapid Adaptation in Reference Tracking of Uncertain Nonlinear Systems21 May 2026
- 2026Physics-Aligned Spectral Mamba: Decoupling Semantics and Dynamics for Few-Shot Hyperspectral Target Detection7 Apr 2026
- 2025Distributionally robust minimization in meta-learning for system identification22 Jun 2025
- 2025Manifold meta-learning for reduced-complexity neural system identification16 Apr 2025
- 2024Enhanced Transformer architecture for in-context learning of dynamical systems4 Oct 2024
- 2024MPC of Uncertain Nonlinear Systems with Meta-Learning for Fast Adaptation of Neural Predictive Models18 Apr 2024
- 2024Learning Neural Differential Algebraic Equations via Operator Splitting19 Mar 2024
- 2023On the adaptation of in-context learners for system identification7 Dec 2023
- 2023From system models to class models: An in-context learning paradigm25 Aug 2023
Similar, not cited
Close in wording, no citation either way
- 2024Deep Learning-based Approaches for State Space Models: A Selective Review
- 2025Meta-Learning for Physically-Constrained Neural System Identification
- 2024Meta-Dynamical State Space Models for Integrative Neural Data Analysis
- 2025The Curious Case of In-Training Compression of State Space Models
- 2024State-space models are accurate and efficient neural operators for dynamical systems
- 2024HiPPO-Prophecy: State-Space Models can Provably Learn Dynamical Systems in Context
- 2024State Space Models as Foundation Models: A Control Theoretic Overview
- 2026MaRK: Markov-adapted Recurrent Kernels for Dynamic Operator Conditioning in State Space Models