HMath.MachineLearning.Preprocessing¶

Description

Fitted preprocessing and index-based splits. Fit on training rows before transforming validation/test rows. Unknown categories are rejected unless explicitly mapped to an all-zero vector. Sample quantiles interpolate between order statistics; constant features use scale one and transform to zero.

On this page¶

API	Purpose
Scaling	Population z-score, unit-range, or median/IQR feature scaling.
Standardize	Declaration and type contract below.
MinMax	Declaration and type contract below.
RobustScale	Declaration and type contract below.
Scaler	Immutable training-feature centers and nonzero scales; constant columns use scale one.
fitScaler	Estimate one center and scale per training feature.
transform	Apply training centers/scales to rows with the original feature count; no state is updated.
inverseTransform	Undo a fitted affine scaler, including its constant-feature convention.
QuantileTransformer	Sorted training samples retained separately for each feature.
fitQuantileTransformer	Fit empirical feature quantiles using training rows only.
quantileTransform	Map values to [0,1].
inverseQuantileTransform	Interpolate training order statistics at probabilities in [0,1].
PowerTransform	Box-Cox or Yeo-Johnson with a caller-selected finite power lambda; lambda is not estimated.
BoxCox	Declaration and type contract below.
YeoJohnson	Declaration and type contract below.
powerTransform	Apply the selected power to features.
inversePowerTransform	Invert the selected power, rejecting values outside the transform range.
Imputation	Training-column mean, median, or a finite constant for NaN-valued observations.
MeanImpute	Declaration and type contract below.
MedianImpute	Declaration and type contract below.
ConstantImpute	Declaration and type contract below.
Imputer	Fitted replacement values; an imputer never learns from rows passed to impute.
fitImputer	Treat NaN as missing and reject infinities.
impute	Replace NaN with fitted training values while preserving observed entries and row order.
Encoder	Sorted distinct training categories with stable zero-based integer codes.
UnknownCategory	Whether a previously unseen category fails or produces an all-zero one-hot row.
RejectUnknown	Declaration and type contract below.
ZeroUnknown	Declaration and type contract below.
fitEncoder	Learn sorted category levels from a nonempty training label vector.
labelEncode	Map known categories to zero-based codes; unseen categories fail.
labelDecode	Recover categories from checked integer codes.
oneHotEncode	Emit one column per sorted training category under the explicit unknown-category policy.
Fold	Disjoint indices into the original dataset.
Fold	Declaration and type contract below.
trainingIndices	Zero-based original row indices assigned to training in this fold.
validationIndices	Zero-based original row indices held out for validation in this fold.
trainTestSplit	Seeded split of n row indices.
kFold	Partition n shuffled indices into k nonempty validation folds; each row is held out exactly once.
stratifiedKFold	Round-robin each shuffled class across k folds.
gridSearch	Minimize a user-supplied validation loss, evaluated independently for every candidate/fold.
randomSearch	Sample count distinct candidates from a finite set, then evaluate the supplied folds.

Documentation

Scaling

data Scaling Source #

Population z-score, unit-range, or median/IQR feature scaling.

Constructors

Standardize
MinMax
RobustScale

Instances

Instances details

Show Scaling Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> Scaling -> ShowS # show :: Scaling -> String # showList :: [Scaling] -> ShowS #
Eq Scaling Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: Scaling -> Scaling -> Bool # (/=) :: Scaling -> Scaling -> Bool #

Scaler

data Scaler Source #

Immutable training-feature centers and nonzero scales; constant columns use scale one.

Instances

Instances details

Show Scaler Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> Scaler -> ShowS # show :: Scaler -> String # showList :: [Scaler] -> ShowS #
Eq Scaler Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: Scaler -> Scaler -> Bool # (/=) :: Scaler -> Scaler -> Bool #

fitScaler

fitScaler :: Scaling -> Dataset -> Either MathError Scaler Source #

Estimate one center and scale per training feature. Empty/ragged or nonfinite data fail.

transform

transform :: Scaler -> Dataset -> Either MathError Dataset Source #

Apply training centers/scales to rows with the original feature count; no state is updated.

inverseTransform

inverseTransform :: Scaler -> Dataset -> Either MathError Dataset Source #

Undo a fitted affine scaler, including its constant-feature convention.

QuantileTransformer

data QuantileTransformer Source #

Sorted training samples retained separately for each feature.

Instances

Instances details

Show QuantileTransformer Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> QuantileTransformer -> ShowS # show :: QuantileTransformer -> String # showList :: [QuantileTransformer] -> ShowS #
Eq QuantileTransformer Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: QuantileTransformer -> QuantileTransformer -> Bool # (/=) :: QuantileTransformer -> QuantileTransformer -> Bool #

fitQuantileTransformer

fitQuantileTransformer :: Dataset -> Either MathError QuantileTransformer Source #

Fit empirical feature quantiles using training rows only.

quantileTransform

quantileTransform :: QuantileTransformer -> Dataset -> Either MathError Dataset Source #

Map values to [0,1]. Ties use their average rank and values outside training extrema clamp.

inverseQuantileTransform

inverseQuantileTransform :: QuantileTransformer -> Dataset -> Either MathError Dataset Source #

Interpolate training order statistics at probabilities in [0,1].

PowerTransform

data PowerTransform Source #

Box-Cox or Yeo-Johnson with a caller-selected finite power lambda; lambda is not estimated.

Constructors

BoxCox Double
YeoJohnson Double

Instances

Instances details

Show PowerTransform Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> PowerTransform -> ShowS # show :: PowerTransform -> String # showList :: [PowerTransform] -> ShowS #
Eq PowerTransform Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: PowerTransform -> PowerTransform -> Bool # (/=) :: PowerTransform -> PowerTransform -> Bool #

powerTransform

powerTransform :: PowerTransform -> Dataset -> Either MathError Dataset Source #

Apply the selected power to features. Box-Cox requires positive values; Yeo-Johnson allows signs.

inversePowerTransform

inversePowerTransform :: PowerTransform -> Dataset -> Either MathError Dataset Source #

Invert the selected power, rejecting values outside the transform range.

Imputation

data Imputation Source #

Training-column mean, median, or a finite constant for NaN-valued observations.

Constructors

MeanImpute
MedianImpute
ConstantImpute Double

Instances

Instances details

Show Imputation Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> Imputation -> ShowS # show :: Imputation -> String # showList :: [Imputation] -> ShowS #
Eq Imputation Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: Imputation -> Imputation -> Bool # (/=) :: Imputation -> Imputation -> Bool #

Imputer

data Imputer Source #

Fitted replacement values; an imputer never learns from rows passed to impute.

Instances

Instances details

Show Imputer Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> Imputer -> ShowS # show :: Imputer -> String # showList :: [Imputer] -> ShowS #
Eq Imputer Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: Imputer -> Imputer -> Bool # (/=) :: Imputer -> Imputer -> Bool #

fitImputer

fitImputer :: Imputation -> Dataset -> Either MathError Imputer Source #

Treat NaN as missing and reject infinities. All-missing columns require ConstantImpute.

impute

impute :: Imputer -> Dataset -> Either MathError Dataset Source #

Replace NaN with fitted training values while preserving observed entries and row order.

Encoder

data Encoder a Source #

Sorted distinct training categories with stable zero-based integer codes.

Instances

Instances details

Show a => Show (Encoder a) Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> Encoder a -> ShowS # show :: Encoder a -> String # showList :: [Encoder a] -> ShowS #
Eq a => Eq (Encoder a) Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: Encoder a -> Encoder a -> Bool # (/=) :: Encoder a -> Encoder a -> Bool #

UnknownCategory

data UnknownCategory Source #

Whether a previously unseen category fails or produces an all-zero one-hot row.

Constructors

RejectUnknown
ZeroUnknown

Instances

Instances details

Show UnknownCategory Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> UnknownCategory -> ShowS # show :: UnknownCategory -> String # showList :: [UnknownCategory] -> ShowS #
Eq UnknownCategory Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: UnknownCategory -> UnknownCategory -> Bool # (/=) :: UnknownCategory -> UnknownCategory -> Bool #

fitEncoder

fitEncoder :: Ord a => [a] -> Either MathError (Encoder a) Source #

Learn sorted category levels from a nonempty training label vector.

labelEncode

labelEncode :: Eq a => Encoder a -> [a] -> Either MathError [Int] Source #

Map known categories to zero-based codes; unseen categories fail.

labelDecode

labelDecode :: Encoder a -> [Int] -> Either MathError [a] Source #

Recover categories from checked integer codes.

oneHotEncode

oneHotEncode :: Eq a => UnknownCategory -> Encoder a -> [a] -> Either MathError Dataset Source #

Emit one column per sorted training category under the explicit unknown-category policy.

Fold

data Fold Source #

Disjoint indices into the original dataset. No estimator is fitted here.

Constructors

Fold
Fields trainingIndices :: [Int] Zero-based original row indices assigned to training in this fold. validationIndices :: [Int] Zero-based original row indices held out for validation in this fold.

Instances

Instances details

Show Fold Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods showsPrec :: Int -> Fold -> ShowS # show :: Fold -> String # showList :: [Fold] -> ShowS #
Eq Fold Source #
Instance details Defined in HMath.MachineLearning.Preprocessing Methods (==) :: Fold -> Fold -> Bool # (/=) :: Fold -> Fold -> Bool #

trainTestSplit

trainTestSplit :: Word64 -> Double -> Int -> Either MathError Fold Source #

Seeded split of n row indices. The fraction denotes test rows, rounded and clamped to [1,n-1].

kFold

kFold :: Word64 -> Int -> Int -> Either MathError [Fold] Source #

Partition n shuffled indices into k nonempty validation folds; each row is held out exactly once.

stratifiedKFold

stratifiedKFold :: Ord a => Word64 -> Int -> [a] -> Either MathError [Fold] Source #

Round-robin each shuffled class across k folds. Every class needs at least k observations.

gridSearch

gridSearch :: [p] -> [Fold] -> (p -> Fold -> Either MathError Double) -> Either MathError (p, Double) Source #

Minimize a user-supplied validation loss, evaluated independently for every candidate/fold. The callback receives only the split indices, so it can fit preprocessing/model state inside each fold. Ties preserve candidate order.

randomSearch

randomSearch :: Word64 -> Int -> [p] -> [Fold] -> (p -> Fold -> Either MathError Double) -> Either MathError (p, Double) Source #

Sample count distinct candidates from a finite set, then evaluate the supplied folds. Lower loss wins.