Knowledge Maxxing← Back to topic library
ML MaxxingBeginner60 sec
Label
A label is the correct answer attached to a training example that a supervised model is trained to predict.
Prerequisite: Training example
What matters
Label quality sets a ceiling on model quality: a model can only ever learn to reproduce the labels it was trained on, including any of their errors or inconsistencies.
Example
In an email dataset, the label for each message is whether a human previously marked it 'spam' or 'not spam'.
Common misconception
A label is not the model's prediction; it is the known correct answer supplied during training, used to judge the model's predictions against.
Recall question
Why does poor label quality limit how good a trained model can be?
Keywords and related concepts
labeltargetground truthfeature