What is machine learning?
Machine learning = get better at a task by learning from data, instead of writing every rule by hand.
Classical programming: you specify the steps. ML: you specify the task, provide data, define how mistakes are scored (loss), and let an algorithm adjust a model.
Handwritten digits
Hand-coding rules for every shape of “8” is hopeless. Collect thousands of labelled images, define a loss (wrong digit = penalty), and train a model that maps pixels → digit.
Not only optimization
Many methods minimize a loss. Others are better seen as statistical inference or pure algorithms. Optimization is still a powerful unifying story.
Key takeaways
- ML learns from data.
- Need task + data + loss.
- Optimization is a common (not exclusive) lens.