banana-bench

A comprehensive collection of 70+ standard mathematical benchmark functions for testing and evaluating optimization algorithms.

Why "banana-bench"? Rosenbrock's function — one of the most famous test functions in optimization, and part of this suite (rosenbrock) — is nicknamed the "banana function" for its curved, banana-shaped valley. This is the bench you run your optimizer through.

📚 Documentation

Full documentation is published at banana-bench.readthedocs.io, built from the sources in the docs/ directory:

• User Guide & Examples: Detailed tutorials on metadata, benchmarking, and utilities.
• Function Reference: Complete guide to all 70+ functions with domains and global minima.
• Visualization Gallery: Examples of all plotting and analysis tools.
• API Reference: Class and function definitions.

🎯 Features

• 70+ Benchmark Functions: Multimodal, Unimodal, and Special functions.
• Rich Metadata: Access bounds, dimensions, known minima programmatically.
• Visualization: 2D/3D plots, convergence tracking, and heatmaps.
• Benchmarking Tools: Automated testing with BenchmarkRunner.
• Zero Core Dependencies: Only NumPy is required.

📦 Installation

From PyPI

pip install banana-bench

From Source

git clone https://github.com/ak-rahul/banana-bench.git
cd banana-bench
pip install -e .

To install with visualization support:

pip install banana-bench[viz]

🚀 Quick Start

import numpy as np
from bananabench import ackley, BenchmarkRunner

# 1. Use a single function
x = np.zeros(5)
print(f"Ackley(0) = {ackley(x)}")

# 2. Run a benchmark suite
def my_optimizer(func, bounds):
    # Your optimization logic here...
    return np.zeros(len(bounds)), 0.0

runner = BenchmarkRunner(my_optimizer, "MyAlgo", n_runs=5)
results = runner.run_suite(functions=['sphere', 'ackley'])

For detailed usage, see the User Guide.