Machine Learning with Probabilistic Law Discovery: A Concise Introduction

TL;DR

This paper introduces Probabilistic Law Discovery (PLD), a logic-based machine learning approach that learns interpretable probabilistic rules for various tasks, emphasizing transparency and applicability across domains.

Contribution

It presents the principles, benefits, limitations, and application guidelines of PLD, a novel rule learning method distinct from decision trees and random forests.

Findings

PLD produces human-readable, interpretable models.

It effectively handles classification, regression, and time series tasks.

PLD offers a transparent alternative to black-box models.

Abstract

Probabilistic Law Discovery (PLD) is a logic based Machine Learning method, which implements a variant of probabilistic rule learning. In several aspects, PLD is close to Decision Tree/Random Forest methods, but it differs significantly in how relevant rules are defined. The learning procedure of PLD solves the optimization problem related to the search for rules (called probabilistic laws), which have a minimal length and relatively high probability. At inference, ensembles of these rules are used for prediction. Probabilistic laws are human-readable and PLD based models are transparent and inherently interpretable. Applications of PLD include classification/clusterization/regression tasks, as well as time series analysis/anomaly detection and adaptive (robotic) control. In this paper, we outline the main principles of PLD, highlight its benefits and limitations and provide some application guidelines.