Machine learning approach to enhance the design of automated theorem provers

  • Authors:

  • Mahdi Khalifa;Hazem Raafat;Mohammed Almulla

  • Affiliations:

  • Computer Science Department, Kuwait University, Safat, Kuwait;Computer Science Department, Kuwait University, Safat, Kuwait;Computer Science Department, Kuwait University, Safat, Kuwait

  • Venue:
  • ICONIP'12 Proceedings of the 19th international conference on Neural Information Processing - Volume Part II
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Extracting knowledge from hard to prove First-order predicate calculus problems and their solutions may lead to improve the performance of automated theorem provers. In this work, we examine the association between the syntactic characteristics of the theorems and the techniques (inference methods and the control strategies) used by various ATPs (automated theorem provers) when attempted these problems. The aim is to determine the most suitable combination of inference methods and control strategies for classes of problems based on their syntactic characteristics. Data mining classification techniques are used to extract knowledge from ATPs attempts on these problems. Such knowledge identifies meta-strategies that are capable to enhance the performance of ATPs. Design improvements in software engineering often enhance the process of software creation, verification, and maintenance. This is one of the primary objectives of this work.