Gomoku AI with Renju Rules 🧠♟️

This is an AI-driven Gomoku (Five in a Row) game implemented with Python 3.13.2 and a PyQt6 graphical interface. It supports human-vs-AI and AI-vs-AI modes under Renju rules, featuring three distinct AI models ranging from classic heuristics to deep reinforcement learning.

✨ Features

• ✅ Renju rule enforcement
• 🎮 PyQt6 GUI for interactive play
• 🤖 Three powerful AI models:
  1. Minimax with heuristic evaluation and Alphabeta pruning
  2. Monte Carlo Tree Search with heuristic rollout
  3. Neural Network (Policy + Value) trained via self-play using MCTS

---

🧠 AI Models

1. Minimax + Heuristic

• Classic tree search
• Static board evaluation heuristics
• Depth-limited for tractability

2. MCTS + Heuristic

• Tree search with playout simulations
• Rollout policy based on simple board heuristics
• Balances exploration & exploitation

3. MCTS + Neural Network (Policy + Value)

• Unified neural network for move probability (policy) and win prediction (value)
• Trained through reinforcement learning from self-play
• Currently trained through 1,000 self played games
• Similar to AlphaZero-style architecture

---

📦 Dependencies

• Python 3.13.2
• NumPy 2.2.2
• PyQt6 6.8.1
• PyTorch 2.6.0

Install them with:

pip install numpy==2.2.2 PyQt6==6.8.1 torch==2.6.0

🚀 Getting Started

python gomoku_gui.py