Core Concepts
Automatic theorem provers benefit from diverse proving strategies in schedules for optimal performance.
Abstract
Automatic theorem provers rely on schedules of diverse proving strategies for optimal performance.
Strategies are discovered and evaluated to construct a schedule for the Vampire prover.
Regularization techniques are proposed to improve schedule performance on unseen problems.
Greedy algorithm is used for schedule construction, outperforming optimal approaches in some cases.
Various regularization methods are explored to enhance the greedy algorithm's performance.
Stats
120 CPU cores were utilized over 20.1 days to cover 6796 problems.
The time limits ranged from 2000 Mi to 256000 Mi during strategy discovery.
The base greedy algorithm found a schedule in less than a minute, while an optimal approach took over 16 hours.
Quotes
"Many tools have been designed to help discover new proving strategies and construct proving schedules."
"Regularization techniques aim at improving the robustness of schedules on unseen data."
"The greedy algorithm can compete with optimal approaches for schedule construction."