Skip to content

API Reference

Overview

This section provides complete technical documentation for the public API of evobench—a standardized benchmarking framework for evolutionary algorithms and metaheuristics. All documented classes and functions are intended for direct use in benchmarking studies.

The reference material is organized into three primary modules:

Core Architecture

The foundation of evobench is built on a strict object-oriented design. All algorithms inherit from a common abstract base class, ensuring:

✓ Uniform interface across all metaheuristics
✓ Consistent state tracking throughout optimization
✓ Standardized initialization protocols
✓ Interchangeable algorithms for fair comparative evaluation

→ Learn more: EvolutionaryAlgorithm Base Class

Algorithms

evobench implements three foundational evolutionary algorithms, each representing distinct solution strategies:

Particle Swarm Optimization (PSO)

Velocity-based swarm intelligence inspired by flocking behavior. Fast convergence with pronounced exploitation in later iterations.

Characteristics: Continuous velocity updates, personal + social learning, good for smooth landscapes
→ Full Reference: PSO Documentation

Estimation of Distribution Algorithm (EDA)

Model-based metaheuristic that learns probabilistic distributions of promising solutions. Robust exploration with systematic sampling.

Characteristics: Gaussian modeling, tournament selection, good for non-separable problems
→ Full Reference: EDA Documentation

Artificial Bee Colony (ABC)

Swarm intelligence inspired by honeybee foraging. Three behavioral roles (employed, onlooker, scout) provide balanced exploration-exploitation.

Characteristics: Multiple bee phases, limit-based abandonment, good for multimodal problems
→ Full Reference: ABC Documentation

Benchmarks

Standardized continuous optimization test functions covering diverse landscape properties:

Function Type Difficulty Best For
Sphere Unimodal Very Low Baseline validation
Rosenbrock Unimodal Medium-High Valley navigation
Ackley Multimodal High Global exploration
Schwefel 1.2 Unimodal Medium Variable coupling
Trid Unimodal High Correlation exploitation

→ Full Reference: Benchmark Functions

Statistics

Automated statistical hypothesis testing framework with decision flow:

  1. Normality Assessment (Shapiro-Wilk test)
  2. Primary Hypothesis Test (ANOVA or Kruskal-Wallis)
  3. Post-Hoc Pairwise Comparisons (Tukey or Dunn's test)

Key Functions: - analyze(): Automatic test selection and execution - stat_report(): Formatted human-readable reports

→ Learn more: See Statistical Theory

Quick API Overview

Import Main Components

# Algorithms
from evobench.algorithms import PSO, EDA, ABC

# Benchmarks
from evobench.benchmarks import sphere, rosenbrock, ackley, schwefel, trid

# Statistical analysis
from evobench.stats import analyze, stat_report

# Base class for custom algorithms
from evobench.base import EvolutionaryAlgorithm

Basic Usage Pattern

from evobench.algorithms import PSO
from evobench.benchmarks import sphere

# Define search domain
bounds = [(-5, 5)] * 10

# Create optimizer
optimizer = PSO(
    objective_function=sphere,
    bounds=bounds,
    population_size=30,
    max_iterations=200
)

# Run optimization
best_solution, best_fitness = optimizer.run()
print(f"Best Fitness: {best_fitness:.6f}")
print(f"Convergence: {optimizer.fitness_history}")

File Structure

reference/
├── index.md                    ← You are here
├── base.md                     # EvolutionaryAlgorithm abstract base
├── benchmarks.md               # Benchmark function registry and API
└── algorithms/
    ├── index.md               # Algorithms overview
    ├── pso.md                 # Particle Swarm Optimization
    ├── eda.md                 # Estimation of Distribution Algorithm
    └── abc.md                 # Artificial Bee Colony

Version Compatibility

This reference documentation applies to evobench 0.1.0+.

For version-specific features and breaking changes, see the Changelog.