Skip to content

Installation

desdeo-brb requires Python 3.10 or later. Install from PyPI:

pip install desdeo-brb          # Core (NumPy + SciPy)
pip install desdeo-brb[jax]     # + JAX backend
pip install desdeo-brb[pyomo]   # + Pyomo/IPOPT backend
pip install desdeo-brb[all]     # Everything

Optional backends

JAX backend

The JAX backend provides JIT-compiled inference and exact-gradient training via automatic differentiation. It is useful on larger problems where finite-difference gradients become a bottleneck.

pip install desdeo-brb[jax]

GPU builds of JAX can be installed separately; see the JAX installation guide for details.

Pyomo / IPOPT backend

The Pyomo backend builds a symbolic optimization model that can be solved with any AMPL-compatible solver. By default the library uses IPOPT.

pip install desdeo-brb[pyomo]

The IPOPT binary must be installed separately:

sudo apt install coinor-libipopt-dev

conda install -c conda-forge ipopt

brew install ipopt

Once the IPOPT binary is on your PATH, you can train with model.fit(X, y, method="ipopt").

Development installation

To contribute or run the test suite, install the dev extras:

git clone https://github.com/gialmisi/desdeo-brb
cd desdeo-brb
uv sync --all-extras
uv run pytest

See CONTRIBUTING for the full contribution guide.

Notebooks

The tutorial notebooks in notebooks/ require jupyter and matplotlib:

pip install desdeo-brb[notebooks]

Then launch with:

jupyter notebook notebooks/