hex.genmodel.algos.deeplearning

Class DeeplearningMojoModel

java.lang.Object

hex.genmodel.GenModel

hex.genmodel.MojoModel

hex.genmodel.algos.deeplearning.DeeplearningMojoModel

All Implemented Interfaces:

IGenModel, java.io.Serializable, IGeneratedModel

public class DeeplearningMojoModel
extends MojoModel

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	DeeplearningMojoModel.StoreWeightsBias

Field Summary

Fields

Modifier and Type	Field and Description
java.lang.String	_activation
double[]	_all_drop_out_ratios
java.lang.String[]	_allActivations
int[]	_catNAFill
int[]	_catoffsets
int	_cats
int[]	_catsA
DistributionFamily	_family
boolean	_imputeMeans
int	_mini_batch_size
double[]	_normmul
double[]	_normrespmul
double[]	_normrespsub
double[]	_normsub
int	_numLayers
int	_nums
double[]	_numsA
int[]	_units
boolean	_use_all_factor_levels
DeeplearningMojoModel.StoreWeightsBias[]	_weightsAndBias

Fields inherited from class hex.genmodel.MojoModel

_balanceClasses, _category, _defaultThreshold, _h2oVersion, _modelClassDistrib, _modelDescriptor, _mojo_version, _nclasses, _nfeatures, _priorClassDistrib, _supervised, _uuid

Fields inherited from class hex.genmodel.GenModel

_domains, _names, _offsetColumn, _responseColumn

Method Summary

Modifier and Type	Method and Description
double	calculateReconstructionErrorPerRowData(double[] original, double[] reconstructed) Calculates average reconstruction error (MSE).
int	getPredsSize(ModelCategory mc)
void	init()
double[]	modifyOutputs(double[] out, double[] preds, double[] dataRow)
double[]	score0(double[] row, double[] preds) Subclasses implement the scoring logic.
double[]	score0(double[] dataRow, double offset, double[] preds) This method will be derived from the scoring/prediction function of deeplearning model itself.

Methods inherited from class hex.genmodel.MojoModel

getModelCategory, getUUID, isSupervised, load, load, nclasses, nfeatures

Methods inherited from class hex.genmodel.GenModel

bitSetContains, bitSetIsInRange, calibrateClassProbabilities, convertDouble2Float, correctProbabilities, createAuxKey, features, GBM_rescale, getColIdx, getDomainValues, getDomainValues, getDomainValues, getHeader, getModelCategories, getNames, getNumClasses, getNumCols, getNumResponseClasses, getPrediction, getPredsSize, getResponseIdx, getResponseName, GLM_identityInv, GLM_inverseInv, GLM_logInv, GLM_logitInv, GLM_ologitInv, GLM_tweedieInv, img2pixels, isAutoEncoder, isClassifier, KMeans_closest, KMeans_distance, KMeans_distance, KMeans_distances, Kmeans_preprocessData, Kmeans_preprocessData, KMeans_simplex, log_rescale, mapEnum, setCats, setCats, setInput, setInput

Methods inherited from class java.lang.Object

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

Field Detail

_mini_batch_size

public int _mini_batch_size

_nums

public int _nums

_cats

public int _cats

_catoffsets

public int[] _catoffsets

_normmul

public double[] _normmul

_normsub

public double[] _normsub

_normrespmul

public double[] _normrespmul

_normrespsub

public double[] _normrespsub

_use_all_factor_levels

public boolean _use_all_factor_levels

_activation

public java.lang.String _activation

_allActivations

public java.lang.String[] _allActivations

_imputeMeans

public boolean _imputeMeans

_units

public int[] _units

_all_drop_out_ratios

public double[] _all_drop_out_ratios

_weightsAndBias

public DeeplearningMojoModel.StoreWeightsBias[] _weightsAndBias

_catNAFill

public int[] _catNAFill

_numLayers

public int _numLayers

_family

public DistributionFamily _family

_numsA

public double[] _numsA

_catsA

public int[] _catsA

Method Detail

init

public void init()

score0

public final double[] score0(double[] dataRow,
                             double offset,
                             double[] preds)

This method will be derived from the scoring/prediction function of deeplearning model itself. However, we followed closely what is being done in deepwater mojo. The variable offset is not used.

Overrides:

score0 in class GenModel

Parameters:

dataRow -

offset -

preds -

Returns:

modifyOutputs

public double[] modifyOutputs(double[] out,
                              double[] preds,
                              double[] dataRow)

score0

public double[] score0(double[] row,
                       double[] preds)

Description copied from class: GenModel

Subclasses implement the scoring logic. The data is pre-loaded into a re-used temp array, in the order the model expects. The predictions are loaded into the re-used temp array, which is also returned. This call exactly matches the hex.Model.score0, but uses the light-weight GenModel class.

Specified by:

score0 in class GenModel

getPredsSize

public int getPredsSize(ModelCategory mc)

Overrides:

getPredsSize in class GenModel

calculateReconstructionErrorPerRowData

public double calculateReconstructionErrorPerRowData(double[] original,
                                                     double[] reconstructed)

Calculates average reconstruction error (MSE). Uses a normalization defined for the numerical features of the trained model.

Returns:

average reconstruction error = ||original - reconstructed||^2 / length(original)