A collection of system identification tools implemented in PyTorch.

• State-space identification methods (see [1], [2], [3], [6])
• Differentiable transfer functions (see [4], [5])

• Examples are provided in the examples folder of this repo.
• The API documentation is available at https://pytorch-ident.readthedocs.io/en/latest.

A Python 3.9 conda environment with

• numpy
• scipy
• matplotlib
• pandas
• pytorch

Run the command

pip install pytorch-ident

This will install the current stable version from the PyPI package repository.

1. Get a local copy the project. For instance, run

git clone https://github.com/forgi86/pytorch-ident.git

in a terminal to clone the project using git. Alternatively, download the zipped project from this link and extract it in a local folder

2. Install pytorch-ident by running

pip install .

in the project root folder (where the file setup.py is located).

[1] M. Forgione and D. Piga. Model structures and fitting criteria for system identification with neural networks. In Proceedings of the 14th IEEE International Conference Application of Information and Communication Technologies, 2020.

[2] B. Mavkov, M. Forgione, D. Piga. Integrated Neural Networks for Nonlinear Continuous-Time System Identification. IEEE Control Systems Letters, 4(4), pp 851-856, 2020.

[3] M. Forgione and D. Piga. Continuous-time system identification with neural networks: model structures and fitting criteria. European Journal of Control, 59:68-81, 2021.

[4] M. Forgione and D. Piga. dynoNet: a neural network architecture for learning dynamical systems. International Journal of Adaptive Control and Signal Processing, 2021.

[5] D. Piga, M.Forgione and M. Mejari. Deep learning with transfer functions: new applications in system identification. In Proceedings of the the 2021 SysId Conference, 2021.

[6] G. Beintema, R. Toth and M. Schoukens. Nonlinear state-space identification using deep encoder networks. Learning for Dynamics and Control. PMLR, 2021.