PySHRED is a deep-learning library for reconstructing and forecasting high-dimensional spatiotemporal systems from sparse sensor data.
Built on the SHallow REcurrent Decoder (SHRED) architecture, PySHRED provides a seamless pipeline from raw sensor measurements to high-fidelity reconstructions and long-horizon forecasts.
| Component | Role | Models |
|---|---|---|
| Sequence model | Encodes temporal sensor measurements into a low-dimensional latent state. | LSTM, GRU, Transformer |
| Decoder model | Reconstructs the full high-dimensional state from the latent state. | MLP, U-Net |
| Latent forecaster | Propagates latent dynamics forward in time for long-horizon prediction. | LSTM, SINDy |
The sequence + decoder pair reconstructs the full high-dimensional state space from sparse sensors, while the forecaster + decoder pair enables multi-step forecasting with no additional sensor measurements.
PySHRED is a powerful tool for:
- System identification
- Reduced-order modeling
- Long-horizon forecasting
- Latent dynamics discovery
- Parametric systems analysis
- Control and decision-making
PySHRED offers a high-level interface and a simple three-step pipeline, making it easy for anyone to get started.
Online documentation: pyshred-dev.github.io/pyshred/stable
The docs include:
-
Installing from PyPI
The latest stable release (and required dependencies) can be installed from PyPI:
pip install pyshred -
Installing from source
PySHRED can be installed via source code on GitHub.
git clone https://github.com/pyshred-dev/pyshred.git cd pyshred pip install .
Citation instructions coming soon.
- Docs: https://pyshred-dev.github.io/pyshred/stable
- Issue Tracking: https://github.com/pyshred-dev/pyshred/issues
- Source code: https://github.com/pyshred-dev/pyshred
|
Nathan Kutz |
Jan Williams |
David Ye |
Mars Gao |
Matteo Tomasetto |
Stefano Riva |
-
Williams, J. P., Zahn, O., & Kutz, J. N. (2024).
Sensing with shallow recurrent decoder networks.
Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 480(2298), 20240054. -
Gao, M. L., Williams, J. P., & Kutz, J. N. (2025).
Sparse identification of nonlinear dynamics and Koopman operators with Shallow Recurrent Decoder Networks.
arXiv preprint, arXiv:2501.13329. -
Tomasetto, M., Williams, J. P., Braghin, F., Manzoni, A., & Kutz, J. N. (2025).
Reduced Order Modeling with Shallow Recurrent Decoder Networks.
arXiv preprint, arXiv:2502.10930. -
Kutz, J. N., Reza, M., Faraji, F., & Knoll, A. (2024).
Shallow Recurrent Decoder for Reduced Order Modeling of Plasma Dynamics.
arXiv preprint, arXiv:2405.11955. -
Ebers, M. R., Williams, J. P., Steele, K. M., & Kutz, J. N. (2024).
Leveraging arbitrary mobile sensor trajectories with Shallow Recurrent Decoder Networks for full-state reconstruction.
IEEE Access, 12, 97428–97439.