rule-learning

Interpretable ML package 🔍 for concise, transparent, and accurate predictive modeling (sklearn-compatible).

Comprehensive suite for rule-based learning

Python library to explain Tree Ensemble models (TE) like XGBoost, using a rule list.

XCSF learning classifier system: rule-based online evolutionary machine learning

The codes for our ACL'22 paper: PRBOOST: Prompt-Based Rule Discovery and Boosting for Interactive Weakly-Supervised Learning.

Learn rule-based models from examples using LLM-powered synthesis. Replace expensive LLM calls with fast, deterministic, inspectable regex, code, or spaCy rules.

A scikit-learn implementation of BOOMER - An Algorithm for Learning Gradient Boosted Multi-label Classification Rules

Rule-Guided Graph Neural Networks for Recommender Systems, ISWC 2020

Implementation of Anticipatory Learning Classifiers System (ALCS) in Python

PyEDCR is a metacognitive neuro-symbolic method for learning error detection and correction rules in deployed ML models using combinatorial sub-modular set optimization

A scikit-learn implementation of BOOMER - An Algorithm for Learning Gradient Boosted Multi-Output Rules

A Java implementation for LORD, a rule learning algorithm proposed in the article "Efficient learning of large sets of locally optimal classification rules" with the approach of searching for a locally optimal rule for each training example. Machine Learning, volume 112, pages 571–610 (2023)

Explain fully connected ReLU neural networks using rules

End-to-end differentiable rule learning for fraud detection. A neural network discovers its own IF-THEN rules via temperature annealing — no hand-coding required.

Research codebase for the project "When rule learning breaks: Diffusion Fails to Learn Parity of Many Bits"

Implementation of pruning hypothesis space using domain theories -- M. Svatoš, G. Šourek, F. Zeležný, S. Schockaert, and O. Kuželka: Pruning Hypothesis Spaces Using Learned Domain Theories, ILP'17

Documentation of the BOOMER machine learning algorithm.

Experimental adaptive policy compiler for AI agents, currently under active development. A hybrid symbolic-LLM engine that turns novel decisions into reusable rules: the LLM handles edge cases and learns new policies, while a fast symbolic runtime executes them cheaply and consistently at scale.

A rule learning algorithm for the deduction of syndrome definitions from time series data.