Neuroevolutionary AI for Games

Bachelor’s Thesis developing an AI through neuroevolution for Magic: The Gathering, combining neural networks with evolutionary algorithms.

Custom Framework

Self-developed system implementing:

• 🧠 Neuroevolutionary algorithms: Neural networks + genetic algorithms
• 🎮 State evaluation: Complex system for game positions
• 📈 Adaptive learning: Evolution across generations
• 📊 Complexity metrics: Quantitative environment analysis

Tech Stack

• Java: Main framework
• Groovy: Automation scripts
• Python: Analysis and evaluation
• Magarena: MTG game engine

Features

• Management of multiple evolutionary generations
• Parallel agent training
• Automated performance evaluation
• Detailed logging and metrics system

Results

The AI demonstrated capability to learn competitive strategies and improve performance generation after generation, validating the applicability of neuroevolutionary methods in highly complex strategic games.