EvoLab

A modern C++23 metaheuristics research platform for solving complex optimization problems.

Overview

EvoLab is a high-performance genetic algorithm framework designed for research-grade optimization of problems like the Traveling Salesman Problem (TSP), Vehicle Routing Problem (VRP), and Quadratic Assignment Problem (QAP).

Key Features

• Modern C++23: Built with the latest C++ features including concepts, ranges, and coroutines
• Research-Grade: Full reproducibility with seed management, comprehensive logging, and statistical analysis
• Hybrid Algorithms: Memetic GA combining genetic operators with local search (2-opt, Lin-Kernighan)
• Advanced Operators: EAX crossover, adaptive operator selection, multiple selection strategies
• High Performance: Optimized data structures, SIMD support, and parallel evaluation
• Extensible Architecture: Template-based design with problem adapters and operator interfaces

Quick Start

Requirements

• C++23 compatible compiler (GCC 12+, Clang 15+, MSVC 19.35+)
• CMake 3.22+
• Optional: oneTBB, OpenMP for parallelization

Building

git clone https://github.com/yourusername/evolab.git
cd evolab
cmake -B build -S . -DCMAKE_BUILD_TYPE=Release
cmake --build build --parallel

Git Hooks (Development)

This project uses lefthook for fast, automated code quality checks:

# Install hooks (one-time setup)
lefthook install

# Hooks now run automatically:
# - Pre-commit: Auto-format code, build check, static analysis
# - Pre-push: Full test suite execution

No manual formatting or linting needed - everything is automated!

Basic Usage

#include <evolab/evolab.hpp>
using namespace evolab;

// Create a TSP problem
auto tsp = problems::create_random_tsp(100);

// Configure and run genetic algorithm
auto ga = factory::make_tsp_ga_basic();
auto result = ga.run(tsp, core::GAConfig{
    .population_size = 256,
    .max_generations = 1000,
    .crossover_prob = 0.9,
    .mutation_prob = 0.1
});

std::cout << "Best solution: " << result.best_fitness.value << std::endl;

Architecture

EvoLab uses a modular, template-based architecture:

• Core: Fundamental concepts, GA implementation, fitness evaluation
• Problems: TSP, VRP, QAP adapters with specialized operators
• Operators: Selection (tournament, roulette), crossover (PMX, OX, EAX), mutation
• Local Search: 2-opt, 3-opt, Lin-Kernighan for memetic algorithms
• Parallel: Thread-pool evaluation, Island Model for distributed evolution

Performance

Target performance on modern hardware:

• 2-opt evaluations: >500M ops/sec (with candidate lists)
• TSP instances: pr2392 (2392 cities) solved to near-optimality in minutes
• Large scale: usa13509 (13509 cities) competitive results in under an hour

Contributing

EvoLab welcomes contributions! Please see CONTRIBUTING.md for guidelines.

Research

If you use EvoLab in research, please cite:

@software{evolab2025,
  title={EvoLab: A Modern C++23 Metaheuristics Platform},
  author={Your Name},
  year={2025},
  url={https://github.com/yourusername/evolab}
}

License

Apache-2.0 - See LICENSE file for details.

Status

🚧 Early Development - Core features implemented, API subject to change.

Roadmap

• Core GA with C++23 concepts
• TSP problem with basic operators
• 2-opt local search
• Advanced crossover (EAX)
• Parallel evaluation
• Island Model
• VRP and QAP problems
• Comprehensive benchmarks
• Python bindings