Normalization
Definition
Normalization brings values to a defined comparable scale.
In ML data preparation, the term depending on the method means scaling to a value range or normalizing a vector. The exact transformation must be specified.
Simply explained
Different sized measuring sticks are brought to a common scale.
Why it matters in practice
Many methods are sensitive to strongly differing magnitudes.
Example from business
Hypothetical: A team scales features before a distance-based classification.
Limits and pitfalls
- Statistics must not learn from later test data. Normalization is not always the same as standardization.
Origin & history
The method originates from statistics, numerical mathematics, and signal processing.
Use in mid-sized companies
- SMEs should document data preparation as a fixed part of their model pipeline.
Frequently asked questions
Is normalization the same as a normalization layer?
No, a layer acts within a network.
Sources
- Preprocessing data — · Scikit-learn: Preprocessing data —
Editorially reviewed on: