Probabilistic grammars for equation discovery
作者：
Highlights：
• Probabilistic grammars outperform deterministic gramamars for equation discovery.
• Parsimony principle parametrized intuitively and flexibly.
• Knowledge expressed through prior distribution over candidates improves performance.
• A simple Monte-Carlo sampling for probabilistic grammar-based equation discovery.
• Computational experiments on 100 benchmark equations confirm theoretical analysis.
摘要
•Probabilistic grammars outperform deterministic gramamars for equation discovery.•Parsimony principle parametrized intuitively and flexibly.•Knowledge expressed through prior distribution over candidates improves performance.•A simple Monte-Carlo sampling for probabilistic grammar-based equation discovery.•Computational experiments on 100 benchmark equations confirm theoretical analysis.
论文关键词：Equation discovery,Symbolic regression,Automated modeling,Grammar,Probabilistic context-free grammar,Monte-Carlo
论文评审过程：Received 30 November 2020, Revised 15 March 2021, Accepted 21 April 2021, Available online 27 April 2021, Version of Record 30 April 2021.
论文官网地址：https://doi.org/10.1016/j.knosys.2021.107077
