hex.genmodel

Class GenModel

java.lang.Object

hex.genmodel.GenModel

All Implemented Interfaces:

IGenModel, java.io.Serializable, IGeneratedModel

Direct Known Subclasses:

MojoModel

public abstract class GenModel
extends java.lang.Object
implements IGenModel, IGeneratedModel, java.io.Serializable

This is a helper class to support Java generated models.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
java.lang.String[][]	_domains Categorical (factor/enum) mappings, per column.
java.lang.String[]	_names Column names; last is response for supervised models
java.lang.String	_offsetColumn Name of the column with offsets (used for certain types of models).

Constructor Summary

Constructors

Constructor and Description
GenModel(java.lang.String[] names, java.lang.String[][] domains)

Method Summary

Methods

Modifier and Type	Method and Description
static boolean	bitSetContains(byte[] bits, int bitoff, double dnum)
static double[]	correctProbabilities(double[] scored, double[] priorClassDist, double[] modelClassDist) Correct a given list of class probabilities produced as a prediction by a model back to prior class distribution
static void	GBM_rescale(double[] preds)
int	getColIdx(java.lang.String name) Returns index of a column with given name, or -1 if the column is not found.
java.lang.String[][]	getDomainValues() Returns domain values for all columns, including the response column.
java.lang.String[]	getDomainValues(int i) Returns domain values for the i-th column.
java.lang.String[]	getDomainValues(java.lang.String name) Gets domain of the given column.
java.lang.String	getHeader() ???
abstract ModelCategory	getModelCategory() Returns this model category.
java.lang.String[]	getNames() The names of all columns used, including response and offset columns.
int	getNumClasses(int colIdx) Get number of classes in the given column.
int	getNumCols() Returns number of columns used as input for training (i.e., exclude response and offset columns).
int	getNumResponseClasses() Return a number of classes in response column.
static int	getPrediction(double[] preds, double[] priorClassDist, double[] data, double threshold) Utility function to get a best prediction from an array of class prediction distribution.
int	getPredsSize() Returns the expected size of preds array which is passed to `predict(double[], double[])` function.
int	getResponseIdx() Returns the index of the response column inside getDomains().
java.lang.String	getResponseName() The name of the response column.
abstract java.lang.String	getUUID() Returns model's unique identifier.
static double	GLM_identityInv(double x)
static double	GLM_inverseInv(double x)
static double	GLM_logInv(double x)
static double	GLM_logitInv(double x)
static double	GLM_tweedieInv(double x, double tweedie_link_power)
boolean	isAutoEncoder() Returns true if this model represents an AutoEncoder.
boolean	isClassifier() Returns true if this model represents a classifier, else it is used for regression.
boolean	isSupervised() Returns true for supervised models.
static int	KMeans_closest(double[][] centers, double[] point, java.lang.String[][] domains, double[] means, double[] mults)
static double	KMeans_distance(double[] center, double[] point, java.lang.String[][] domains, double[] means, double[] mults)
static double	KMeans_distance(double[] center, float[] point, java.lang.String[][] domains, double[] means, double[] mults, double[] colSum, double[] colSumSq)
static double[]	KMeans_simplex(double[][] centers, double[] point, java.lang.String[][] domains, double[] means, double[] mults)
static double	log_rescale(double[] preds)
double[]	map(java.util.Map<java.lang.String,java.lang.Double> row, double[] data) Takes a HashMap mapping column names to doubles.
int	mapEnum(int colIdx, java.lang.String enumValue) Maps given column's categorical to the integer used by this model (returns -1 if mapping not found).
int	nclasses() Returns number of output classes for classifiers, 1 for regression models, and 0 for unsupervised models.
int	nfeatures() Returns number of input features.
float[]	predict(double[] data, float[] preds) Predict the given row and return prediction.
float[]	predict(double[] data, float[] preds, int maxIters) Predict the given row and return prediction using given number of iterations (e.g., number of trees from forest).
abstract double[]	score0(double[] row, double[] preds) Subclasses implement the scoring logic.
double[]	score0(double[] row, double offset, double[] preds)
double[]	score0(java.util.Map<java.lang.String,java.lang.Double> row)
double[]	score0(java.util.Map<java.lang.String,java.lang.Double> row, double[] preds)
double[]	score0(java.util.Map<java.lang.String,java.lang.Double> row, double[] data, double[] preds)

Methods inherited from class java.lang.Object

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

Field Detail

_names

public final java.lang.String[] _names

Column names; last is response for supervised models

_domains

public final java.lang.String[][] _domains

Categorical (factor/enum) mappings, per column. Null for non-enum cols. Columns match the post-init cleanup columns. The last column holds the response col enums for SupervisedModels.

_offsetColumn

public java.lang.String _offsetColumn

Name of the column with offsets (used for certain types of models).

Constructor Detail

GenModel

public GenModel(java.lang.String[] names,
        java.lang.String[][] domains)

Method Detail

isSupervised

public boolean isSupervised()

Returns true for supervised models.

Specified by:

isSupervised in interface IGenModel

Returns:

true if this class represents supervised model.

nfeatures

public int nfeatures()

Returns number of input features.

Specified by:

nfeatures in interface IGenModel

Returns:

number of input features used for training.

nclasses

public int nclasses()

Returns number of output classes for classifiers, 1 for regression models, and 0 for unsupervised models.

Specified by:

nclasses in interface IGenModel

Returns:

returns number of output classes or 1 for regression models.

getModelCategory

public abstract ModelCategory getModelCategory()

Returns this model category.

Specified by:

getModelCategory in interface IGenModel

Returns:

model category

See Also:

ModelCategory

getUUID

public abstract java.lang.String getUUID()

Description copied from interface: IGeneratedModel

Returns model's unique identifier.

Specified by:

getUUID in interface IGeneratedModel

getNumCols

public int getNumCols()

Returns number of columns used as input for training (i.e., exclude response and offset columns).

Specified by:

getNumCols in interface IGeneratedModel

getNames

public java.lang.String[] getNames()

The names of all columns used, including response and offset columns.

Specified by:

getNames in interface IGeneratedModel

getResponseName

public java.lang.String getResponseName()

The name of the response column.

Specified by:

getResponseName in interface IGeneratedModel

getResponseIdx

public int getResponseIdx()

Returns the index of the response column inside getDomains().

Specified by:

getResponseIdx in interface IGeneratedModel

getNumClasses

public int getNumClasses(int colIdx)

Get number of classes in the given column. Return number greater than zero if the column is categorical or -1 if the column is numeric.

Specified by:

getNumClasses in interface IGeneratedModel

getNumResponseClasses

public int getNumResponseClasses()

Return a number of classes in response column.

Specified by:

getNumResponseClasses in interface IGeneratedModel

Returns:

number of response classes

isClassifier

public boolean isClassifier()

Returns true if this model represents a classifier, else it is used for regression.

Specified by:

isClassifier in interface IGeneratedModel

Returns:

true if this model represents a classifier, else it is used for regression.

isAutoEncoder

public boolean isAutoEncoder()

Returns true if this model represents an AutoEncoder.

Specified by:

isAutoEncoder in interface IGeneratedModel

Returns:

true if this model represents an AutoEncoder.

getDomainValues

public java.lang.String[] getDomainValues(java.lang.String name)

Gets domain of the given column.

Specified by:

getDomainValues in interface IGeneratedModel

Parameters:

name - column name

Returns:

return domain for given column or null if column is numeric.

getDomainValues

public java.lang.String[] getDomainValues(int i)

Returns domain values for the i-th column.

Specified by:

getDomainValues in interface IGeneratedModel

Parameters:

i - index of column

Returns:

domain for given categorical column or null if columns contains numeric value

getDomainValues

public java.lang.String[][] getDomainValues()

Returns domain values for all columns, including the response column.

Specified by:

getDomainValues in interface IGeneratedModel

getColIdx

public int getColIdx(java.lang.String name)

Returns index of a column with given name, or -1 if the column is not found.

Specified by:

getColIdx in interface IGeneratedModel

mapEnum

public int mapEnum(int colIdx,
          java.lang.String enumValue)

Maps given column's categorical to the integer used by this model (returns -1 if mapping not found).

Specified by:

mapEnum in interface IGeneratedModel

getPredsSize

public int getPredsSize()

Returns the expected size of preds array which is passed to `predict(double[], double[])` function.

Specified by:

getPredsSize in interface IGeneratedModel

Returns:

expected size of preds array

predict

public float[] predict(double[] data,
              float[] preds)

Description copied from interface: IGeneratedModel

Predict the given row and return prediction.

Specified by:

predict in interface IGeneratedModel

Parameters:

data - row holding the data. Ordering should follow ordering of columns returned by getNames()

preds - allocated array to hold a prediction

Returns:

returned preds parameter filled by prediction

predict

public float[] predict(double[] data,
              float[] preds,
              int maxIters)

Description copied from interface: IGeneratedModel

Predict the given row and return prediction using given number of iterations (e.g., number of trees from forest).

Specified by:

predict in interface IGeneratedModel

Parameters:

data - row holding the data. Ordering should follow ordering of columns returned by getNames()

preds - allocated array to hold a prediction

maxIters - maximum number of iterations to use during predicting process

Returns:

returned preds parameter filled by prediction

map

public double[] map(java.util.Map<java.lang.String,java.lang.Double> row,
           double[] data)

Takes a HashMap mapping column names to doubles.

Looks up the column names needed by the model, and places the doubles into the data array in the order needed by the model. Missing columns use NaN.

score0

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

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.

score0

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

score0

public double[] score0(java.util.Map<java.lang.String,java.lang.Double> row,
              double[] data,
              double[] preds)

score0

public double[] score0(java.util.Map<java.lang.String,java.lang.Double> row,
              double[] preds)

score0

public double[] score0(java.util.Map<java.lang.String,java.lang.Double> row)

correctProbabilities

public static double[] correctProbabilities(double[] scored,
                            double[] priorClassDist,
                            double[] modelClassDist)

Correct a given list of class probabilities produced as a prediction by a model back to prior class distribution

The implementation is based on Eq. (27) in the paper.

Parameters:

scored - list of class probabilities beginning at index 1

priorClassDist - original class distribution

modelClassDist - class distribution used for model building (e.g., data was oversampled)

Returns:

corrected list of probabilities

getPrediction

public static int getPrediction(double[] preds,
                double[] priorClassDist,
                double[] data,
                double threshold)

Utility function to get a best prediction from an array of class prediction distribution. It returns the index of the max. probability (if that exists). In the case of ties, it samples from the tied classes with the likelihood given by the prior probabilities.

Parameters:

preds - an array of prediction distribution. Length of arrays is equal to a number of classes+1.

priorClassDist - prior class probabilities (used to break ties)

data - Test data

threshold - threshold for binary classifier

Returns:

the best prediction (index of class, zero-based)

bitSetContains

public static boolean bitSetContains(byte[] bits,
                     int bitoff,
                     double dnum)

KMeans_closest

public static int KMeans_closest(double[][] centers,
                 double[] point,
                 java.lang.String[][] domains,
                 double[] means,
                 double[] mults)

KMeans_simplex

public static double[] KMeans_simplex(double[][] centers,
                      double[] point,
                      java.lang.String[][] domains,
                      double[] means,
                      double[] mults)

KMeans_distance

public static double KMeans_distance(double[] center,
                     float[] point,
                     java.lang.String[][] domains,
                     double[] means,
                     double[] mults,
                     double[] colSum,
                     double[] colSumSq)

KMeans_distance

public static double KMeans_distance(double[] center,
                     double[] point,
                     java.lang.String[][] domains,
                     double[] means,
                     double[] mults)

log_rescale

public static double log_rescale(double[] preds)

GBM_rescale

public static void GBM_rescale(double[] preds)

GLM_identityInv

public static double GLM_identityInv(double x)

GLM_logitInv

public static double GLM_logitInv(double x)

GLM_logInv

public static double GLM_logInv(double x)

GLM_inverseInv

public static double GLM_inverseInv(double x)

GLM_tweedieInv

public static double GLM_tweedieInv(double x,
                    double tweedie_link_power)

getHeader

public java.lang.String getHeader()

???