What Is Machine Learning?
In traditional programming, you write explicit rules. In Machine Learning, the model learns rules from examples. If the data is good and the model is trained well, it can generalize to unseen inputs.
We learn parameters $\theta$ so predictions $\hat{y}$ are close to true targets $y$.
Simple Mental Model
Data teaches patterns, loss measures error, optimization reduces loss, and evaluation checks if learning generalizes.
Major Types of ML
| Type | Data | Goal | Example |
|---|---|---|---|
| Supervised | Labeled $(x, y)$ | Predict known targets | Spam detection, house prices |
| Unsupervised | Unlabeled $x$ | Find hidden structure | Customer segmentation |
| Reinforcement | Rewards from environment | Learn best actions over time | Game-playing agents |
Typical ML Workflow
1. Define the problem and success metric
2. Collect and clean data
3. Split data (train / validation / test)
4. Choose baseline model
5. Train and optimize (reduce loss)
6. Evaluate on unseen data
7. Iterate and deployStrong baselines and clean evaluation usually matter more than complex models early on.
Core Math Intuition
Chapter 1 starts here: how we minimize $\mathcal{L}(\theta)$ efficiently using Gradient Descent, SGD, Momentum, and Adam.
Before You Start Chapter 1
Quick Readiness Checklist
Be comfortable with basic algebra, plotting graphs, and the idea of derivatives as slopes. You do not need advanced math to begin.