How Machine Learning Works

Machine learning (ML) is the branch of AI that enables systems to learn from data rather than being explicitly programmed with rules. Instead of telling a computer how to solve a problem step by step, you show it many examples and let it discover the patterns itself.

---

Traditional Programming vs Machine Learning

Traditional Programming:
┌──────────────┐     ┌───────────┐     ┌──────────┐
│    Rules     │────▶│  Program  │────▶│  Output  │
│    Data      │────▶│           │     │          │
└──────────────┘     └───────────┘     └──────────┘

Machine Learning:
┌──────────────┐     ┌───────────┐     ┌──────────┐
│    Data      │────▶│ Learning  │────▶│  Rules   │
│   Outputs    │────▶│ Algorithm │     │ (Model)  │
└──────────────┘     └───────────┘     └──────────┘

In traditional programming, a developer writes explicit rules. In machine learning, you provide examples, and the algorithm learns the patterns. The output is a model — a mathematical function that can make predictions on new, unseen data.

---

Key Concepts and Terminology

Term	Definition
Features	The input variables used to make predictions
Labels	The correct answers in the training data
Training data	The dataset used to teach the model
Test data	A separate dataset for evaluation
Model	The mathematical function learned from data
Prediction	The model's output for new input
Hyperparameters	Configuration settings chosen before training
Epoch	One complete pass through the training dataset

Train/Test Split

Total Dataset (e.g., 10,000 samples)
├── Training Set (80%) ─── 8,000 samples ─── Used to train the model
├── Validation Set (10%) ─ 1,000 samples ─── Used to tune hyperparameters
└── Test Set (10%) ─────── 1,000 samples ─── Used for final evaluation only

Important: The test set must never be used during training. Using it would give an overly optimistic estimate of performance.

---

Types of Machine Learning

1. Supervised Learning

The model learns from labelled examples — input-output pairs where the correct answer is known.

Regression predicts a continuous value (house prices, blood pressure, sales revenue). Classification predicts a discrete category (spam/not spam, dog breed, disease type).

Common algorithms: linear regression, logistic regression, decision trees, random forests, SVMs, k-nearest neighbours.

2. Unsupervised Learning

No labels. The algorithm finds structure in the data on its own.

Clustering groups similar data points (customer segmentation, anomaly detection). Dimensionality reduction simplifies data while preserving key information (PCA, t-SNE).

3. Reinforcement Learning

An agent learns by interacting with an environment, receiving rewards for good actions and penalties for bad ones.