Documentation

Additional notes on the experiments can be found under the notes folder.

Installation

The virtual environment can be installed with the command uv sync. Ensure that it gets created under .venv in the root directory as pyright is configured to check for this folder.

Datasets

For our experiments we use the TU datasets. The dataset is downloaded on the fly, however it is usually convenient to preprocess the data before training the model. The datasets can be downloaded and preprocessed by running the following command in the terminal.

uv run datasets

Training

To train a model using the configuration provided, run the train script with a provided configuration which can be found under the config folder. One can train a single model with the following command.

uv run src/train.py configs/nervepool/nervepool-DD-2020.yaml

The sweep-*.yaml files are used to generate the set of configs for the experiments in each of the subfolders. To create a set of configs with such a sweep file, we run the command uv run sweep path/to/sweep-file.yaml.

The training script will create a similar named file in the results directory and are under version control. To create the tables, run the script tables.py under the notebooks folder.