MetaHeurOpt

Description

MetaHeurOpt is a modular Python library designed for the execution, benchmarking, and statistical analysis of metaheuristic optimization algorithms. It provides a standardized framework to compare diverse algorithms, ranging from classic evolutionary strategies to modern physics-inspired and swarm intelligence models, on complex continuous optimization problems.

Installation

You can install the stable version of MetaHeurOpt directly from PyPI:

pip install metaheuropt

Usage Example

Below is a quick start guide to comparing solvers. For a more detailed walkthrough, including statistical tables and 3D landscape visualization, check out the /examples/example.ipynb notebook.

import numpy as np
from metaheuropt.solvers import GA, PSO
from metaheuropt.core import Optimizer
from metaheuropt.analysis import ResultsAnalyzer

# 1. Define your objective function
def sphere(x):
    return np.sum(np.square(x))

# 2. Configure experiment
dims = 10
bounds = (np.array([-5.12] * dims), np.array([5.12] * dims))
num_runs = 5
max_iter = 100

# 3. Select solvers
solvers = [
    GA(bounds=bounds, pop_size=50, max_iter=max_iter),
    PSO(bounds=bounds, pop_size=50, max_iter=max_iter),
]

# 4. Run optimization
opt = Optimizer(solvers=solvers, obj_func=sphere, num_runs=num_runs)
opt.run(save_results=True, results_folder="demo_results")

# 5. Analyze and visualize
analyzer = ResultsAnalyzer("demo_results")
analyzer.plot_convergence()
analyzer.perform_stats(full=True)

Available Algorithms

The library currently supports the following high-performance metaheuristics: