Induction of Non-Monotonic Rules From Statistical Learning Models Using High-Utility Itemset Mining
This provides a faster, more scalable method for explainable AI in inductive logic programming, though it appears incremental as it builds on existing HUIM and SHAP techniques.
The authors tackled the problem of inducing non-monotonic logic programs from statistical learning models by reducing clause search to High-Utility Itemset Mining (HUIM) and using TreeExplainer for feature importance extraction. Their experiments on UCI benchmarks showed significant improvements in classification metrics and training time compared to the state-of-the-art ALEPH system.
We present a fast and scalable algorithm to induce non-monotonic logic programs from statistical learning models. We reduce the problem of search for best clauses to instances of the High-Utility Itemset Mining (HUIM) problem. In the HUIM problem, feature values and their importance are treated as transactions and utilities respectively. We make use of TreeExplainer, a fast and scalable implementation of the Explainable AI tool SHAP, to extract locally important features and their weights from ensemble tree models. Our experiments with UCI standard benchmarks suggest a significant improvement in terms of classification evaluation metrics and running time of the training algorithm compared to ALEPH, a state-of-the-art Inductive Logic Programming (ILP) system.