Classical Machine Learning

Welcome to Lesson 6 of Artificial Intelligence Concepts (COMP5511). This session serves as a bridge from theoretical foundations to practical, algorithmic implementations. While modern AI often emphasizes Deep Learning, Classical Machine Learning remains the bedrock of data analytics. These algorithms offer high interpretability and computational efficiency, making them the preferred choice for structured data and industry-standard analytics.

1. Supervised Learning

This paradigm involves training a model on a labeled dataset, where the algorithm learns the relationship between input features and a specific target output. This allows the model to predict outcomes for new, unseen data accurately.

Decision Trees: Models that split data into branches to reach a classification or numerical decision.
Support Vector Machines (SVMs): Algorithms that find the optimal hyperplane to maximize the margin between different data classes.

2. Unsupervised Learning

These algorithms analyze unlabeled data to discover hidden patterns, structures, or groupings without any prior guidance on what the output should be. Key techniques include:

K-means Clustering: Grouping data points into K-distinct groups based on feature similarities.
Principal Component Analysis (PCA): A dimensionality reduction technique used to simplify complex data while retaining its essential variance.

Interpretability vs. Complexity

A significant advantage of classical machine learning is its transparency. Unlike "black-box" deep learning models, algorithms like Decision Trees allow humans to trace the exact logic behind a prediction, which is vital for high-stakes fields like medicine or finance.

Scikit-learn implementation workflow