Machine learning
머신러닝
Also known as: ML
Instead of writing the rules, you let a system find them in data.
To recognize a cat, should you write down the shape of its ears? Machine learning shows the system many cat photos instead and lets it find what they share.
Learning comes in three broad kinds: supervised learning from labelled data, unsupervised learning that finds structure without labels, and reinforcement learning that rewards good actions.
Data decides the result. A skewed dataset produces skewed behaviour, which is why data review matters most in products used by children.
- Versus rule based codeUse it where the rules cannot be written by hand.
- What it needsData and an evaluation standard come before the model.