pyensmallen: python bindings for the ensmallen library for numerical optimization

Lightweight python bindings for ensmallen library. Currently supports

• L-BFGS, with intended use for optimisation of smooth objectives for m-estimation
• ADAM (and variants with different step-size routines) - makes use of ensmallen's templatization.
• Frank-Wolfe, with intended use for constrained optimization of smooth losses
  • constraints are either lp-ball (lasso, ridge, elastic-net) or simplex
• (Generalized) Method of Moments estimation with ensmallen optimizers.
  • this uses ensmallen for optimization [and relies on jax for automatic differentiation to get gradients and jacobians]. This is the main use case for pyensmallen and is the reason for the bindings.
• Estimator classes for linear, logistic, and Poisson regression with classical and robust inference for unregularized fits

See ensmallen docs for details. The notebooks/ directory walks through several statistical examples.

speed

pyensmallen is very fast. A comprehensive set of benchmarks is available in the benchmarks directory. The benchmarks are run on an intel 12th gen framework laptop. Benchmarks vary data size (sample size and number of covariates) and parametric family (linear, logistic, poisson) and compare pyensmallen with scipy and statsmodels (I initially also tried to keep cvxpy in the comparison set but it was far too slow to be in the running). At large data sizes, pyensmallen is roughly an order of magnitude faster than scipy, which in turn is an order of magnitude faster than statsmodels. So, a single statsmodels run takes around as long as a pyensmallen run that naively uses the nonparametric bootstrap for inference. This makes the bootstrap a viable option for inference in large data settings.

Installation:

Make sure your system has blas installed. On macos, this can be done via brew. Linux systems should have it installed by default. If you are using conda, you can install blas via conda-forge.

Then,

from pypi

uv pip install pyensmallen

from source

1. Install armadillo and ensmallen for your system (build from source, or via conda-forge; I went with the latter)
2. git clone this repository
3. If you are using uv:
   • uv pip install --python .venv/bin/python meson meson-python ninja pybind11
   • uv pip install --python .venv/bin/python --no-build-isolation -e .
4. If you are using vanilla pip in an activated environment:
   • python -m pip install meson meson-python ninja pybind11
   • python -m pip install --no-build-isolation -e .
5. Profit? Or at least minimize loss?

full development environment

To install everything required to run tests and notebooks:

uv pip install --python .venv/bin/python meson meson-python ninja pybind11
uv pip install --python .venv/bin/python --no-build-isolation -e ".[full]"

Vanilla pip equivalent:

python -m pip install meson meson-python ninja pybind11
python -m pip install --no-build-isolation -e ".[full]"

The full extra includes the Python dependencies used by:

• the test suite
• GMM and autodiff examples
• benchmark notebooks
• plotting and notebook tooling

documentation

doc-generation

The repository includes a Quarto documentation site in docs/. The docs are built from three sources:

• hand-written Quarto pages in docs/*.qmd
• generated API reference pages in docs/reference/*.qmd, built from Python and pybind11 docstrings with quartodoc
• executed notebook pages in docs/notebooks/*.ipynb

Use the render script instead of calling quarto render directly:

scripts/render_docs.sh

The script does the following:

• uses the repository .venv as the Quarto Python runtime
• forces JAX onto CPU so notebook execution is stable during docs builds
• copies the tracked notebooks from notebooks/ into docs/notebooks/
• runs quartodoc to regenerate the API reference pages from docstrings
• runs quarto render docs to execute the notebooks and build the site

If you need the full docs toolchain first:

uv pip install --python .venv/bin/python meson meson-python ninja pybind11
uv pip install --python .venv/bin/python --no-build-isolation -e ".[full]"

The rendered site lands in docs/_site/. The generated API source pages land in docs/reference/.

from wheel

• download the appropriate .whl for your system from the more recent release listed in Releases and run uv pip install ./pyensmallen... OR
• copy the download url and run uv pip install https://github.com/apoorvalal/pyensmallen/releases/download/<version>/pyensmallen-<version>-<pyversion>-linux_x86_64.whl

Documentation

GitHub Pages is served from the gh-pages branch. This repository does not auto-publish the website when master changes, so updating the site is a separate step.

After rendering the documentation site into docs/_site/, publish it with:

scripts/publish_docs.sh

The script:

• creates a temporary worktree for gh-pages
• syncs docs/_site/ into that worktree
• commits the site update if anything changed
• pushes gh-pages to origin