a Preprocessing that converts an Array[Double] or Array[Float] to a Tensor.
Convert a BigDL Transformer to a Preprocessing
chains two Preprocessing together.
construct a Preprocessing that convert (Feature, Label) tuple to a Sample.
construct a Preprocessing that convert (Feature, Label) tuple to a Sample. The returned Preprocessing is robust for the case label = None, in which the Sample is derived from Feature only.
data type from feature column, E.g. Array[_] or Vector
data type from label column, E.g. Float, Double, Array[_] or Vector
Preprocessing defines data transform action during feature preprocessing.
Preprocessing defines data transform action during feature preprocessing. Multiple Preprocessing can be combined into a ChainedPreprocessing. E.g., FeatureStep1[A, B] -> FeatureStep2[B, C] yield a ChainedFeatureSteps[A, C]
input data type
output data type
It represents the relationship between two items.
A relation pair is made up of two relations of the same id1: Relation(id1, id2Positive, label>0) [Positive Relation] Relation(id1, id2Negative, label=0) [Negative Relation]
converts numbers to Tensors.
a Preprocessing that converts Float, Double, Array[Float], Array[Double] or MLlib Vector to a Tensor.
a Preprocessing that converts Tensor to Sample.
a Preprocessing that converts MLlib Vector to a Tensor.
a Preprocessing that converts MLlib Vector to a Tensor. :: deprecated, NNEstimator can automatically extract Vectors now.
(Since version 0.4.0) NNEstimator can automatically extract Vectors now
chains two Preprocessing together. The output type of the first Preprocessing should be the same with the input type of the second Preprocessing.
input type of the first Preprocessing
output type of the first Preprocessing, as well as the input type of the last Preprocessing
output of the last Preprocessing