Symbolic Regression Genetic Algorithm

This project implements a genetic algorithm in Go to perform symbolic regression. It evolves mathematical expressions to fit datasets by mimicking biological evolution processes.

Features

• Expression Representation: Uses binary trees to represent and evolve mathematical functions.
• Speciation: Maintains population diversity by grouping similar individuals, preventing premature convergence.
• Elitism: Ensures the best-performing individuals are carried over to the next generation.
• Genetic Operators: Includes mutation and crossover to explore the solution space.
• Dataset Management: Logic to generate synthetic datasets or load existing ones.
• History Export: Evolution progress and statistics are exported as JSON for post-run analysis.

Project Structure

• main.go: Entry point for dataset generation, evolution execution, and JSON export.
• analysis.ipynb: A Jupyter Notebook designed to read the JSON history files and visualize the results.

Usage

1. Run Evolution:

   go run main.go datagen.go config.json ga_config.json

   or use the pre-compiled binary:

   ./symbolic_regression config.json ga_config.json

2. Analyze Results: Open analysis.ipynb in a Jupyter environment to load the generated JSON data and inspect fitness curves and evolved expressions.

Credits

• Inspired by work of Kenneth Stanley phd thesis, paper