Module normalize 

Per-feature min-max / z-score normalizer.

Rescales a D-dimensional point into [0, 1] (MinMax) or (value − μ) / σ (ZScore) given per-dimension NormParams learned from a batch or loaded from a saved baseline. The None strategy is an identity transform kept so the normalizer can be inserted into a detection pipeline even before fit data arrives.

Policy-free — the lib stores params and applies the math; caller decides when to refit, when to swap, and what the source of the samples is.

Structs

NormParams

Per-dimension learned parameters.

Normalizer

Fixed-dim normalizer with per-feature parameters.

Enums

NormStrategy

Per-dim transform strategy.