# Query-driven PAC-Learning for Reasoning

## TL;DR

This paper introduces a method for learning reasoning rules from data within the PAC-Semantics framework, enabling proof-supporting rule learning during backward proof search in standard logics.

## Contribution

It demonstrates how backward proof search algorithms can be adapted to learn rules supporting proofs, applicable to logics like chaining and resolution.

## Key findings

- Algorithms can learn rules during proof search under PAC-Semantics.
- Applicable to standard logics such as chaining and resolution.
- Enhances rule learning in logical reasoning systems.

## Abstract

We consider the problem of learning rules from a data set that support a proof of a given query, under Valiant's PAC-Semantics. We show how any backward proof search algorithm that is sufficiently oblivious to the contents of its knowledge base can be modified to learn such rules while it searches for a proof using those rules. We note that this gives such algorithms for standard logics such as chaining and resolution.