Knowledge Maxxing← Back to topic library
ML MaxxingBeginner60 sec
Dataset
A dataset is the full collection of training examples gathered for a machine learning task, before it is split into training, validation and test portions.
Prerequisite: Training example
What matters
A dataset's coverage and quality determine what patterns a model can possibly learn; a model cannot learn a pattern that never appears in its dataset.
Example
A dataset for a fraud model might contain millions of past transactions, each with its features and a label of whether it was later confirmed fraudulent.
Common misconception
A larger dataset does not automatically fix a biased or unrepresentative sample; more of the wrong kind of data does not add missing coverage.
Recall question
Why can't a model learn a pattern that never appears anywhere in its dataset?
Keywords and related concepts
datasetdata collectiontraining examplestraining set