hex.genmodel.easy.prediction

Class AutoEncoderModelPrediction

java.lang.Object

hex.genmodel.easy.prediction.AbstractPrediction

hex.genmodel.easy.prediction.AutoEncoderModelPrediction

All Implemented Interfaces:

java.io.Serializable

public class AutoEncoderModelPrediction
extends AbstractPrediction

Data reconstructed by the AutoEncoder model based on a given input.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
double	mse Reconstruction mean squared error calculated from original and reconstructed data.
double[]	original Representation of the original input the way AutoEncoder model sees it (1-hot encoded categorical values)
double[]	reconstructed Reconstructed data, the array has same length as the original input.
RowData	reconstructedRowData Reconstructed data represented in RowData structure.

Constructor Summary

Constructors

Constructor and Description
AutoEncoderModelPrediction()

Method Summary

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

original

public double[] original

Representation of the original input the way AutoEncoder model sees it (1-hot encoded categorical values)

reconstructed

public double[] reconstructed

Reconstructed data, the array has same length as the original input. The user can use the original input and reconstructed output to easily calculate eg. the reconstruction error.

reconstructedRowData

public RowData reconstructedRowData

Reconstructed data represented in RowData structure. The structure will copy the structure of the RowData input with the exception of categorical values. Categorical fields will be represented as a map of the domain values to the reconstructed values. Example: input RowData([sex: "Male", ..]) will produce output RowData([sex: [Male: 0.9, Female: 0.1], ..]

mse

public double mse

Reconstruction mean squared error calculated from original and reconstructed data. Uses a normalization defined for the numerical features of the trained model. average reconstruction error = ||original - reconstructed||^2 / length(original)

Constructor Detail

AutoEncoderModelPrediction

public AutoEncoderModelPrediction()