mononet 0.0.0a0

Unconstrained monotonic neural networks for PyTorch, JAX, and Keras

mononet — Unconstrained Monotonic Neural Networks

Reference implementation of the unconstrained monotonic neural network construction from:

Runje, D., Shankaranarayana, S. M. (2023). Constrained Monotonic Neural Networks. ICML 2023. https://arxiv.org/abs/2205.11775

First-class support for PyTorch, JAX (Flax NNX), and Keras 3.

Install

pip install "mononet[torch]"      # PyTorch
pip install "mononet[jax]"        # JAX + Flax NNX
pip install "mononet[keras]"      # Keras 3
pip install "mononet[all]"        # all three

Quick start

A 60-second tour will appear here once the algorithm implementation lands. Each backend exposes the same composed model (MonoMLP) and the framework-idiomatic layer name (MonoLinear for PyTorch and JAX, MonoDense for Keras).

# PyTorch
from mononet.torch import MonoMLP

# JAX
from mononet.jax import MonoMLP

# Keras 3
from mononet.keras import MonoMLP

License

Apache License 2.0 — see LICENSE and NOTICE.md. Commercial use is permitted. The technique is described in U.S. Patent 11,551,063 (assignee: AIRT Technologies Ltd.); the Apache-2.0 license grants the patent rights needed to use this code. For academic use, please cite the paper (see NOTICE.md).

Formal proofs

Every theorem in the paper is mechanized in Lean 4 + mathlib4 under proofs/. See the cross-reference page for the paper-claim ↔ Lean-theorem ↔ Python-test mapping.

Documentation

Full docs at https://davorrunje.github.io/mononet/. Source for guides and benchmarks lives in docs/docs/.

Contributing

See CONTRIBUTING.md for the development workflow: devcontainer choice, uv sync, pre-commit, per-backend test commands.

Citation

If you use mononet in academic work, please cite the paper. BibTeX is in docs/docs/about/citation.md.