Introduction to Machine Learning

What Is Machine Learning?

The key word in that definition is teaches itself. No programmer writes the rules explicitly — the algorithm finds them in the data.

In traditional software, a developer writes rules and the computer applies them. In supervised machine learning, the developer provides data and labeled outputs, and the algorithm learns a function that maps inputs to predictions. In unsupervised learning, there are no labels — the algorithm finds structure on its own.

This course covers how machines learn and the three main paradigms. It pairs with Gradient Descent & Optimization and Overfitting & Regularization in this track.

The Three Learning Paradigms

Supervised Learning

In supervised learning, the model is trained on labeled examples — input-output pairs where the correct answer is known. The model learns to map inputs to outputs by minimising the error between its predictions and the true labels.

• Classification: predict a category. For e.g., spam or not spam, cat or dog.
• Regression: predict a continuous value. For e.g., house price or temperature forecast.
• Common algorithms: Linear Regression, Logistic Regression, Decision Trees, Support Vector Machines, Neural Networks.

Unsupervised Learning

Unsupervised learning discovers structure in unlabeled data. There are no correct answers provided — the algorithm finds patterns on its own.

• Clustering: group similar data points together. For e.g., K-Means grouping customers by purchase behaviour.
• Dimensionality reduction: compress data to fewer dimensions while preserving structure. For e.g., PCA and t-SNE.
• Generative models: learn the data distribution to generate new samples. For e.g., VAEs and GANs for image synthesis. Whether these belong under unsupervised or their own generative category is debated — the field has not settled on a single classification.

Reinforcement Learning

Reinforcement learning trains an agent to make sequential decisions in an environment. The agent receives a reward signal after each action and learns a policy that maximises long-term reward.

• Key components: Agent, Environment, State, Action, Reward, Policy.
• Famous applications: AlphaZero (pure self-play RL), ChatGPT fine-tuned via RLHF, robot locomotion. Note: the original AlphaGo combined supervised learning from human game records with RL — it was not pure reinforcement learning.
• Core algorithms: Q-Learning, PPO, A3C.

The Bias-Variance Tradeoff

A model with high bias is too simple — it fails to capture the true pattern and performs poorly on both training and new data. A model with high variance is too complex — it fits training data perfectly but fails on unseen examples. The goal is a model complex enough to capture true patterns but not so complex that it fits noise.

Key Concepts

• Training set: data used to fit model parameters.
• Validation set: data used to tune hyperparameters and catch overfitting during development.
• Test set: held-out data used once at the end to estimate real-world performance.
• Hyperparameters: configuration values set before training. For e.g., learning rate, number of layers, regularization strength.
• Cross-validation: estimate generalisation by rotating which portion of data is held out.
• Feature engineering: creating or transforming input variables to improve model performance.

Classification Metrics

Accuracy alone can be misleading. For e.g., a classifier that always predicts "not fraud" on a dataset where only 5% of cases are fraud will score 95% accuracy while catching zero fraud cases.

• Precision: of all predicted positives, what fraction are truly positive? TP / (TP + FP).
• Recall: of all actual positives, what fraction did the model catch? TP / (TP + FN).
• F1 Score: harmonic mean of precision and recall — balances both.
• ROC-AUC: area under the ROC curve, measuring overall discrimination ability.