GESR: A Genetic Programming-Based Symbolic Regression Method with Gene Editing
Researchers have introduced GESR, a novel symbolic regression approach that enhances traditional Genetic Programming (GP) by integrating gene editing techniques. While standard GP relies on random mutations and crossovers to evolve mathematical formulas from data, this randomness often leads to inefficient exploration of solution spaces. To address this, the proposed method employs two BERT models, metaphorically termed the 'hands of God,' to guide the evolutionary process. One model utilizes masked language modeling to direct gene mutations, while the other predicts optimal crossover points. This targeted approach aims to eliminate detrimental variations and accelerate the discovery of beneficial mathematical laws. Experimental results indicate that GESR significantly improves computational efficiency compared to classical GP algorithms. Furthermore, the method demonstrates strong overall performance across various symbolic regression tasks, offering a more effective tool for discovering underlying natural phenomena through artificial intelligence. This development represents a significant advancement in combining large language model capabilities with evolutionary algorithms for scientific discovery.
Wire timeline
GESR: A Genetic Programming-Based Symbolic Regression Method with Gene Editing
Researchers have introduced GESR, a novel symbolic regression approach that enhances traditional Genetic Programming (GP) by integrating gene editing techniques. While standard GP relies on random mutations and crossovers to evolve mathematical formulas from data, this randomness often leads to inefficient exploration of solution spaces. To address this, the proposed method employs two BERT models, metaphorically termed the 'hands of God,' to guide the evolutionary process. One model utilizes masked language modeling to direct gene mutations, while the other predicts optimal crossover points. This targeted approach aims to eliminate detrimental variations and accelerate the discovery of beneficial mathematical laws. Experimental results indicate that GESR significantly improves computational efficiency compared to classical GP algorithms. Furthermore, the method demonstrates strong overall performance across various symbolic regression tasks, offering a more effective tool for discovering underlying natural phenomena through artificial intelligence. This development represents a significant advancement in combining large language model capabilities with evolutionary algorithms for scientific discovery.
cs.AI updates on arXiv.org