Loglinear models for first-order probabilistic reasoning

TL;DR

This paper introduces a loglinear model framework for first-order probabilistic reasoning using stochastic logic programs, enabling probabilistic inference directly on proofs and integrating ILP for feature induction from data.

Contribution

It presents a novel approach combining loglinear models with stochastic logic programs for first-order reasoning, including a method for feature induction via ILP.

Findings

Probabilities are defined on proofs and atomic formulae.

Framework extends first-order reasoning with probabilistic semantics.

ILP can induce features for the loglinear model from data.

Abstract

Recent work on loglinear models in probabilistic constraint logic programming is applied to first-order probabilistic reasoning. Probabilities are defined directly on the proofs of atomic formulae, and by marginalisation on the atomic formulae themselves. We use Stochastic Logic Programs (SLPs) composed of labelled and unlabelled definite clauses to define the proof probabilities. We have a conservative extension of first-order reasoning, so that, for example, there is a one-one mapping between logical and random variables. We show how, in this framework, Inductive Logic Programming (ILP) can be used to induce the features of a loglinear model from data. We also compare the presented framework with other approaches to first-order probabilistic reasoning.