COIT29224 Evolution Strategy Optimizer

Overview

This repository provides a modular Python implementation of an Evolution Strategy (ES) algorithm for optimizing complex, multi-modal objective functions. The project is designed for Assignment 2 of the COIT29224 Evolutionary Computation course at CQUniversity, and demonstrates key ES concepts including:

• Population-based stochastic search
• Variation through Gaussian mutation
• Configurable survival selection: (μ, λ) and (μ + λ)

The codebase emphasizes clarity, extensibility, and robust documentation.

Key Features

• Modular OOP Design: Core logic is encapsulated in EvolutionStrategy and Individual classes.
• Selection Strategies: Supports both (μ, λ) (comma) and (μ + λ) (plus) survivor selection.
• Gaussian Mutation: Configurable mutation strength (sigma) for exploration.
• Boundary Handling: Solutions are clipped to remain within user-specified bounds.
• Benchmark Functions: Includes the Rastrigin function; easily extensible for others.
• Visualization: Generates convergence plots and, for 2D problems, a landscape plot showing the best solution.
• Logging: Each run is logged with a unique batch ID, and plots are saved in batch-specific folders.
• Reproducibility: Supports random seeds for repeatable experiments.
• Extensible: Add new objective functions by editing a single file.

Project Structure

.
├── LICENSE                         # Project license (GPL)
├── README.md                       # This documentation file
├── es_optimiser/                   # Main Python package for ES logic
│   ├── __init__.py                 # Package initializer
│   ├── __pycache__/                # Python bytecode cache
│   ├── evolution_strategy.py       # ES algorithm implementation
│   ├── objective_functions.py      # Fitness/objective functions (e.g., Rastrigin)
│   └── plot.py                     # Plotting utilities
├── logs/                           # Output logs directory
├── main.py                         # Main script to run the optimizer
├── plan-to-develop.md              # Future development notes
├── requirements.txt                # Python package dependencies
├── plots/                          # Output plots directory
│   ├── <batch_id>/                 # Run-specific plots
│   └── plots.md                    # Plot documentation/notes
└── report.md                       # Project report

See ES Algorithm Flowchart (diagram.md)

Installation

Requirements:

• Python 3.8 or newer
• NumPy
• Matplotlib

Setup:

1. Clone the repository:

   git clone https://github.com/Romerolweb/COIT29224-EvolutionStrategy.git
   cd COIT29224-EvolutionStrategy

2. (Recommended) Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: .\venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

How to Run

Run the main script from the project root:

python main.py

What happens:

• The script prints the ES configuration and progress.
• Upon completion, it prints the best solution found.
• Convergence and (if 2D) landscape plots are saved in a batch-specific folder under plots/.
• All run details are logged in the logs/ directory.

Configuration

Edit the parameters in main.py (inside the run_es_optimization function or the main block):

• PROBLEM_DIMENSIONS: Number of variables (int)
• SEARCH_BOUNDS: Tuple of (min, max) for each variable
• POPULATION_MU: Number of parents (μ)
• OFFSPRING_LAMBDA: Number of offspring (λ)
• GENERATIONS: Number of generations
• MUTATION_SIGMA: Mutation strength (σ)
• SELECTION_STRATEGY: '(mu, lambda)' or '(mu + lambda)'
• RANDOM_SEED: Integer or None
• objective_function: Function to optimize (e.g., rastrigin)

Adding Objective Functions

1. Define your function in es_optimiser/objective_functions.py:

   def my_function(x: np.ndarray) -> float:
       # Your implementation
       return value

2. Update main.py to use your function:

   from es_optimiser.objective_functions import my_function
   # ...
   es_optimizer = EvolutionStrategy(
       objective_function=my_function,
       # ...
   )

Output

• Logs: Each run creates a log file in logs/ with a unique batch ID.
• Plots: Convergence and landscape plots are saved in plots/<batch_id>/.
• Console: Progress and summary statistics are printed.

License

This project is for academic use. See LICENSE.

Author

• Sebastian Romero Laguna
• Orcid: