AI Course Labs

Repository with projects and assignments from my Artificial Intelligence course, covering fundamental AI techniques from classical machine learning to evolutionary computation and fuzzy systems.

---

Overview

This repository contains 10 lab assignments that explore core concepts in Artificial Intelligence and Machine Learning, progressing from foundational data manipulation through classical ML, deep learning, evolutionary algorithms, and fuzzy logic systems.

#	Topic
1	NumPy & Matplotlib
2	Data Preprocessing
3	Decision Trees
4	PyTorch & Activation Functions
5	MLP for Binary Addition
6	Classification Models
7	Computer Vision & CNNs
8	Evolutionary Computation
9	Aircraft Design Optimization
10	Fuzzy Logic Systems

---

Concepts Covered

Data Preprocessing & Visualization

Feature engineering fundamentals for ML pipelines. Covers standardization (removing mean, scaling to unit variance), Min-Max scaling, MaxAbs scaling, L2 normalization of samples, and encoding categorical features. Uses NumPy for array operations (creation, indexing, slicing, broadcasting, random sampling) and Matplotlib for data visualization.

Classical Machine Learning

Supervised learning with Decision Trees on the Iris dataset. Includes data cleaning, label encoding, train/test splitting, entropy-based tree construction, hyperparameter tuning via GridSearchCV, and tree visualization.

Neural Networks & Deep Learning

Progressive exploration of deep learning with PyTorch:

• Autograd : automatic differentiation, gradient computation for activation functions (Sigmoid, ReLU, Linear)
• Multilayer Perceptrons : feedforward networks for binary addition (XOR-like problem), comparing ReLU, Tanh, and Sigmoid activations
• Classification : linear vs. non-linear classifiers, multiclass classification pipeline
• Computer Vision : image tensors, convolution kernels (blur, sharpen, edge detection), pooling, training CNNs vs. linear models

Evolutionary Computation

Bio-inspired optimization techniques:

• Genetic Algorithms : selection, crossover, mutation with binary representation
• Multi-Objective Optimization : NSGA-II with non-dominated sorting, crowding distance, Pareto front analysis
• Real-World Application : aircraft design optimization with 6 continuous variables (wing span, area, sweep angle, thickness-to-chord ratio, fuselage length/diameter), constraint handling via penalty functions, SBX crossover, polynomial mutation

Fuzzy Logic & Rule-Based Systems

Reasoning under uncertainty using Mamdani fuzzy inference:

• Membership functions : triangular and trapezoidal
• Fuzzification : mapping crisp inputs to fuzzy membership degrees
• Linguistic rules : IF–THEN rules with min for implication and max for accumulation
• Defuzzification : centroid method for crisp output
• Case study : autonomous delivery robot speed recommendation based on visibility, obstacle distance, and battery level