Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

1.0.1 - 2026-06-19

Added

  • py.typed marker so type information ships with the package and is picked up by downstream type checkers (PEP 561), matching the Typing :: Typed classifier.
  • Community health files: GitHub issue and pull request templates.

Fixed

  • Corrected the Zenodo DOI in the README badge to match CITATION.cff.
  • CONTRIBUTING.md now references the correct default branch (master) and the docstring style actually used in the codebase (Google, matching mkdocs.yml).

Dependencies

  • Bumped several development, docs, and notebook dependencies via Dependabot (e.g. jupyterlab, notebook, jupyter-server, tornado, bleach, mistune, urllib3, idna, pymdown-extensions). No changes to the core runtime dependencies.

1.0.0 - 2026-04-16

Added

  • Initial stable release of desdeo-brb 1.x.x, a trainable Belief Rule-Based inference system implementing the RIMER methodology (Yang et al. 2006; Chen et al. 2011).
  • Core BRBModel class with scikit-learn-compatible fit() and predict() API.
  • NumPy backend with SLSQP and trust-constr optimizers for standard MSE training.
  • JAX backend with L-BFGS-B and automatic differentiation for fast training of large models.
  • Pyomo/IPOPT backend for use with custom symbolic objectives.
  • Differential Evolution (DE) and hybrid DE+SLSQP training methods for non-convex problems.
  • Multi-start optimization via n_restarts parameter to handle local minima.
  • Adaptive referential value training as described in Chen et al. (2011).
  • Explainability features: describe_rule(), describe_all_rules(), InferenceResult.explain(), and BRBModel.explain() for human-readable rule descriptions and prediction traces.
  • Custom loss function support via fit_custom() for domain-specific objectives such as INFRINGER-style value function learning.
  • Four Jupyter notebooks covering getting started, multi-attribute models, expert knowledge integration with pipeline leak detection, and explainability.

Dependencies

  • Core: numpy>=1.24, scipy>=1.10, pydantic>=2.0
  • Optional: jax (for JAX backend), pyomo (for IPOPT backend), jupyter + matplotlib (for running the notebooks)