Bayesian tensor-decomposition model for epidemiological surveillance data, demonstrated on Project Tycho weekly case counts. The model decomposes disease incidence into location, type, and temporal effects using Gaussian process priors with FFT-accelerated inference via NumPyro.
Make sure you have uv installed and execute the following commands to install dependencies and the package.
uv sync --all-groups
uv pip install -e .Verify the installation by running the tests.
uv run pytest tests/Run the full analysis pipeline (download data, fit model, generate results):
make allOr run individual steps:
make data # Download and preprocess Project Tycho dataset
make fit # Fit the model (uses configs/helmert.py by default)
make results # Generate visualization notebookConfiguration and output paths can be customized:
make fit CONFIG=configs/default.py OUTPUT=workspace/default