WarpTails#
- class sdm.processing.numerical.WarpTails(*, probability: float = 0.4, exponent: tuple[float, float] = (0.05, 5.0), scale: tuple[float, float] = (0.3, 2.0))#
Bases:
ProcessorWarp numerical columns to more skew or heavier tails.
When the processor is fitted, each column is selected with
probability. A selected column gets one of two maps: \(x^{a}\) for \(x \geq 0\) and \(-|x|^{b}\) otherwise, which adds skew; or \(\mathrm{sign}(x) (e^{s |x|} - 1)\), which adds heavy tails. Standardize the columns before this step.
Capabilities#
Handled Semantic Types |
|
Requires Fitting |
✅ |