Feature scaling
Feature scaling: Transforming feature scales to make distances, gradients and numerical contributions comparable during training or analysis. Standardisation and min-max serve different needs. Statistics must be estimated on training data to avoid leakage into validation and test. This perspective connects the definition with practical behaviour in artificial intelligence systems.
In italiano: Feature ScalingLegge I · Embedding, feature e spazi vettorialiStandardisation and min-max serve different needs. Statistics must be estimated on training data to avoid leakage into validation and test.