Learning from data rather than rules.
THE DEFINITION
Systems that learn patterns from examples rather than following rules a programmer wrote.
HOW IT DIFFERS FROM ORDINARY SOFTWARE
Ordinary software follows instructions. Machine learning derives them from data.
WHY THAT MATTERS
Some problems have rules too numerous or subtle to write.
Distinguishing spam from legitimate email is one. Recognising a face is another.
THE THREE MAIN TYPES
- Supervised: learning from labelled examples
- Unsupervised: finding structure without labels
- Reinforcement: learning from outcomes
WHAT IT REQUIRES
Data, in quantity, relevant to the problem.
That is usually the constraint rather than the technique.
WHAT IT PRODUCES
A model: a set of learned values turning input into output.
Frequently without any explanation of why.
RELATED TERMS
Training, model, deep learning.
