Scholarlink

This paper improves SAT solvers by optimizing the initialization of truth values for variables in a CNF formula. It introduces a Multilayer Perceptron (MLP) model that predicts optimal Genetic Algorithm (GA) parameters based on formula characteristics. This method automates parameter tuning through machine learning, enhancing efficiency and reducing manual adjustments. Experimental results show that our MLP-guided GA approach improves SAT solver performance, indicating a promising direction for automating SAT initialization.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving SAT Solver Performance Through MLP-Predicted Genetic Algorithm Parameters

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sabrine Saouli,
  • Souheib Baarir,
  • Claude Dutheillet

摘要

This paper improves SAT solvers by optimizing the initialization of truth values for variables in a CNF formula. It introduces a Multilayer Perceptron (MLP) model that predicts optimal Genetic Algorithm (GA) parameters based on formula characteristics. This method automates parameter tuning through machine learning, enhancing efficiency and reducing manual adjustments. Experimental results show that our MLP-guided GA approach improves SAT solver performance, indicating a promising direction for automating SAT initialization.