Robust ML for CPS

Repository for the paper "Robustness and Generalization Performance of Deep Learning Models on Cyber-Physical Systems: A Comparative Study" by Alexander WIndmann, Henrik Steude and Oliver Niggemann.

Get started

Install the dependencies

From the root of this repo, run:

poetry install

Generate the data

The data is generated by running '1-generate-data.ipynb' in the notebooks folder. The notebook will also provide visualizations to analyze the datasets.

Run the model training

From the root of this repo, do:

unset LD_LIBRARY_PATH
poetry run python run_training.py

Run the model fine-tuning

From the root of this repo, do:

unset LD_LIBRARY_PATH
poetry run python run_finetuning.py  --LOG_DIR logs/<your-path>

Folder structure

---

├── README.md          <- The top-level README for developers using this project.
│
├── data               <- Datasets and lightning datamodules.
│   └── processed      <- The final, canonical data sets for modeling.
│
├── logs               <- Trained and serialized models, model predictions, or model summaries.
│   
├── models             <- Lightning modules of all the models.
│
├── notebooks          <- Jupyter notebooks.
│
└── visualizations     <- Generated graphics and figures to be used in reporting.

---