Adaptive Search-Space Scaling (AdSS) for Genetic Algorithms

This repository contains the source code, experimental scripts, and visualisation results associated with the study on adaptive search-space scaling for genetic algorithms and self-configuring evolutionary algorithms.

The project investigates a cluster-guided dynamic modification of the search domain based on the Individual Cluster Efficiency (ICE) criterion. The proposed approach is evaluated both for a classical binary-coded Genetic Algorithm (GA) and for a Self-Configuring Genetic Algorithm (SelfCGA), demonstrating improved convergence speed, solution quality, and robustness on multimodal benchmark problems.

---

📁 Repository Contents

📊 Experimental Results (PDF)

• D10_curves_ga_.pdf — Convergence curves for GA and GA with adaptive search-space scaling at (D=10)

• GA_vs_ClusterGA_whiskers_D10_all.pdf — Boxplots for GA and cluster-guided GA at (D=10)

• D10_curves_selfcga_.pdf — Convergence curves for SelfCGA and SelfCGA with scaling at (D=10)

• selfcga_vs_ClusterSelfcga_whiskers_D10_all_.pdf — Boxplots for SelfCGA and its scaled variant at (D=10)

• D30_curves_selfcga_.pdf — Convergence curves for SelfCGA and scaled SelfCGA at (D=30)

• selfcga_vs_ClusterSelfcga_whiskers_D30_all_.pdf — Boxplots at (D=30)

• D50_curves_selfcga_.pdf — Convergence curves for SelfCGA and scaled SelfCGA at (D=50)

• selfcga_vs_ClusterSelfcga_whiskers_D50_all_.pdf — Boxplots at (D=50)

---

🧠 Notebooks

• GA.ipynb — Baseline classical Genetic Algorithm
• GA + mode.ipynb — GA with adaptive search-space scaling (cluster-guided)
• SelfCGA.ipynb — Baseline self-configuring genetic algorithm
• SelfCGA + mode.ipynb — SelfCGA with adaptive search-space scaling

---

⚙️ Requirements

To run the experiments, the following external dependencies are required:

1. CEC 2017 Benchmark Suite (Python implementation)

Required for all test functions:

D. T. <CEC 2017 implementation> (official benchmark package)

The functions are expected to be available under:

from cec2017 import basic, transforms

2. thefittest Python library

Full description on https://github.com/sherstpasha/thefittest. Required for the SelfCGA implementation:

pip install thefittest

3. Standard scientific Python stack

numpy
scipy
scikit-learn
matplotlib