Definition
Machine Learning (ML) is a field of artificial intelligence (AI) where computers learn patterns from data and improve performance without being explicitly programmed.
- Instead of writing rules by hand, we give the system examples (data).
- The model learns relationships between inputs (features) and outputs (labels).
- Once trained, it can make predictions on new, unseen data.
Core Idea
$\text{Model: } y = f(X) + \epsilon$
- $X$ = input features
- $y$ = output (prediction)
- $f(\cdot)$ = function learned from data
- $\epsilon$ = error/noise
Types of Machine Learning
- Supervised Learning
- Learn from labeled data (input → correct output).
- Tasks:
- Regression → predict continuous value (house price, temperature).
- Classification → predict categories (spam/not spam, disease/healthy).
- Unsupervised Learning
- Learn from unlabeled data (no given output).
- Tasks:
- Clustering → group similar data (customer segmentation).
- Dimensionality Reduction → compress features (PCA, embeddings).
- Semi-Supervised Learning
- Mix of labeled + unlabeled data.
- Common when labels are expensive (medical images).
- Reinforcement Learning (RL)
- Agent learns by interacting with environment and receiving rewards.
- Example: AlphaGo, robotics, recommendation systems.
- Self-Supervised Learning (modern deep learning trend)
- Model learns from raw data by predicting part of the input from other parts (e.g., masked words in BERT).
- Core idea behind LLMs like GPT.
Applications
- Computer Vision: image classification, object detection, medical imaging.
- Natural Language Processing (NLP): chatbots, translation, summarization.
- Healthcare: disease prediction, drug discovery.
- Finance: fraud detection, algorithmic trading.
- E-commerce: recommendation systems.
- Autonomous Systems: self-driving cars, robotics.
Workflow in ML
- Collect data
- Clean & preprocess
- Choose model (linear regression, decision tree, neural net, etc.)
- Train model (optimize weights)
- Evaluate on test data
- Deploy to production
- Monitor & retrain (continuous learning)
Example
- Input: house size (1000 sqft), number of rooms (3).
- Output: price = $250,000.
- Train ML model on thousands of such examples.
- Model learns: price increases with size & rooms.
- Later → given new input (1200 sqft, 4 rooms), model predicts price ≈ $300,000.
Why It Matters
- Automates pattern discovery.
- Handles massive, complex datasets humans can’t.
- Powers modern AI systems (LLMs, vision models, recommender systems).
Summary
Machine Learning = teaching computers to learn patterns from data and make predictions without explicit programming.
- Main types: supervised, unsupervised, reinforcement.
- Applications: vision, NLP, healthcare, finance, e-commerce.
