Machine learning: finding patterns in examples — Computing, 11–13 years
In machine learning, people give a computer many examples instead of writing every rule by hand. The computer finds patterns in those examples and uses them to make a useful guess.
Learning from examples
A machine-learning system studies examples that people provide, such as pictures labelled “cat” or “not cat”. It adjusts its internal pattern until it can make a guess about a new picture. It does not learn like a person; it finds useful regularities in data.
Why learn from examples?
Some tasks are too messy to describe with a short list of rules. Recognising speech, faces or spam depends on many small clues that are hard to write down. Examples let a computer use those clues without a person spelling out every detail.
Sorting messages
A filter sees 100 messages, 60 marked “ordinary” and 40 marked “spam”. Step 1: it measures clues such as repeated adverts or strange links. Step 2: it compares a new message with the patterns. Step 3: it predicts “spam”, but the prediction can still be wrong.
A prediction is not certainty
It is easy to think that a system that learned from examples must always know the answer. That is reasonable because its confident results can look clever. But it can meet a new case, a poor example or an unfair pattern, so its answer needs checking.
Recommendations and accessibility
Machine learning can suggest a song, turn spoken words into captions or help a camera identify objects. These uses can be helpful, but they work from past examples and may fail for people or situations unlike those examples. Human judgement still matters.
Keep exploring
Other languages
Loading MyLeoNes™…