Skip to content

Inria-Chile/ces-global-convergence

Repository files navigation

Global Convergence of the Canonical Evolutionary Strategy

This repository contains the official implementation and numerical experiments for the paper: "From Mean-Field Limits to Semiclassical Concentration: Global Convergence of the Canonical Evolutionary Strategy" (2026).

📋 Overview

This project implements the Canonical Evolutionary Strategy (CES), a global optimization framework. We provide the tools to reproduce the theoretical validation and the benchmark results presented in the paper.

Key Features:

  • CES Engine: Core implementation of the evolutionary dynamics.
  • Mean-Field Limits: Numerical verification of the $M^{-1/2}$ convergence rate.
  • Semiclassical Analysis: Scripts for visualizing the concentration of the ground state.
  • High-Dimensional Benchmarks: Scalability tests in $d=1$ to $d=30$ for Ackley function at different initialization regimes (Uniform and Shifted).

📁 Repository Structure

  • /ces_engine.py: Main algorithm logic.
  • /plot_theory_figures.py: Mathematical validation plots.
  • /plot_mass_transport.py: 1D and 2D mass transport visualizations.
  • /run_benchmarks.py: High-performance simulation suite.
  • /data/: (Stored in Parquet) Raw simulation results.

🛠 Installation & Usage

  1. Clone the repo: git clone https://github.com/inria-chile/ces-global-convergence.git
  2. Install dependencies: pip install numpy pandas matplotlib scipy pyarrow
  3. Run benchmarks: python run_benchmarks.py

🎓 Citation

If you use this code in your research, please cite our paper:

@misc{ces2026global,
  title={From Mean-Field Limits to Semiclassical Concentration: Global Convergence of the Canonical Evolutionary Strategy},
  author={Neto, Mat{\'\i}as and Garay, Nicolas and Mart{\'\i}, Luis and Sanchez-Pi, Nayat},
  journal={arXiv preprint},
  note={Submitted to PPSN 2026}
  year={2026},
}

About

Official implementation of the Canonical Evolutionary Strategy (CES). This project provides a reproducible pipeline to verify theoretical mean-field convergence rates and global optimization performance on high-dimensional benchmarks.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors