symbolic-equation

Symbolic regression discovers scientific equations from data using LLM-guided evolutionary search. It balances exploration and exploitation through multi-island evolution and softmax-based cluster sampling to extract interpretable models from physical datasets.

Yes, you can use LLMs to find scientific models in physics and biology. The evolutionary search uses LLM-proposed equation mutations to mine interpretable equations applicable across these scientific domains.

To start symbolic regression, provide a dataset description, list your variables, and define the target. This initializes the evolutionary search to explore the equation space and optimize parameters for your specific data.

The best way to balance exploration and exploitation is using softmax-based cluster sampling within a multi-island evolutionary search. This mechanism efficiently navigates the equation space while LLMs propose targeted mutations.

Yes, the evolutionary algorithm optimizes parameters during fitness evaluation. It calculates fitness using negative mean squared error combined with parameter optimization to ensure the discovered equations accurately fit the data.