Adaptive Systems Programming Practice: Exploring Intelligent Algorithms from Cellular Automata to Neural Networks

This project is a collection of experiments from the Adaptive Systems Programming course at the Autonomous University of Nuevo León (UANL) in Mexico, covering topics such as automatic regulation systems, cellular automata, fuzzy logic systems, and neural networks, with Python and Java implementations. It provides abundant practical cases for learning intelligent systems and computational intelligence. Its core value lies in the combination of theory and practice, helping learners gradually deepen their understanding from basic rules to complex intelligent algorithms.

Adaptive systems are a core direction in AI and computational intelligence, covering technologies from rule evolution to neural network learning. This project has a clear structure, divided into four main modules (automatic regulation, cellular automata, fuzzy systems, neural networks) and six additional practical exercises. Each module includes theory and programming implementation, and some experiments have assignment evidence documents, supporting step-by-step learning by topic.

Key content of each module:

• Automatic Regulation Systems: Includes sentinel light controller (PID regulation), automatic irrigation system (multi-input decision-making), and adaptive traffic mode selector (multi-objective optimization), demonstrating the perception-decision-execution closed-loop architecture.
• Cellular Automata: 1D binary implementation (Python, Wolfram rules) and Java version, embodying how local rules generate global complex patterns.
• Fuzzy Systems: Implements Mario Kart track classification using the skfuzzy library, covering the processes of fuzzification, rule base construction, inference, and defuzzification.
• Neural Networks: MLP (Parkinson's dataset), CNN (MNIST recognition), K-means clustering (native and sklearn implementations), covering supervised and unsupervised learning.

Additional exercises include ant colony optimization (TSP problem), Flood It game (tkinter), firework particle simulation (pygame), K-means clustering (classic datasets), complex network analysis, and agent & reinforcement learning (Q-learning maze solving). The technology stack covers Python (NumPy/Pandas, scikit-learn, TensorFlow/Keras, etc.), Java, Jupyter Notebook, etc., supporting multi-language comparison and diverse tool applications.

This project is suitable for computer science students (to consolidate theory), AI beginners (to build a knowledge chain), self-learners (abundant examples and documents), educators (course materials), and transitioning engineers (to understand algorithm details). Its features include equal emphasis on theory and practice, progressive difficulty, multi-language comparison, real application scenarios, and team collaboration elements.

Project limitations: Insufficient error handling in some code, documents mainly in Spanish, and outdated APIs for deep learning frameworks. Improvement suggestions: Add English documents, add unit tests, update dependency libraries, and supplement visual outputs.

This project is a rich and well-structured learning resource for intelligent systems, covering topics from classic AI to modern machine learning. Learners can master the implementation of various intelligent algorithms and understand the core ideas of adaptive systems (from local rules to collective intelligence, from programming to data-driven). It is recommended for students, engineers, and AI enthusiasts, emphasizing the importance of hands-on practice.