org.deeplearning4j.eval

Class ROCBinary

java.lang.Object

org.deeplearning4j.eval.BaseEvaluation<ROCBinary>

org.deeplearning4j.eval.ROCBinary

All Implemented Interfaces:

Serializable, IEvaluation<ROCBinary>

public class ROCBinary
extends BaseEvaluation<ROCBinary>

ROC (Receiver Operating Characteristic) for multi-task binary classifiers. As per ROC, ROCBinary supports both exact (thersholdSteps == 0) and thresholded; see ROC for details.

Unlike ROC (which supports a single binary label (as a single column probability, or 2 column 'softmax' probability distribution), ROCBinary assumes that all outputs are independent binary variables. This also differs from ROCMultiClass, which should be used for multi-class (single non-binary) cases.

ROCBinary supports per-example and per-output masking: for per-output masking, any particular output may be absent (mask value 0) and hence won't be included in the calculated ROC.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
static int	DEFAULT_STATS_PRECISION

Constructor Summary

Constructors

Constructor and Description
ROCBinary()
ROCBinary(int thresholdSteps)
ROCBinary(int thresholdSteps, boolean rocRemoveRedundantPts)

Method Summary

Modifier and Type	Method and Description
double	calculateAUC(int outputNum) Calculate the AUC - Area Under (ROC) Curve Utilizes trapezoidal integration internally
double	calculateAverageAuc() Macro-average AUC for all outcomes
void	eval(org.nd4j.linalg.api.ndarray.INDArray labels, org.nd4j.linalg.api.ndarray.INDArray networkPredictions)
void	eval(org.nd4j.linalg.api.ndarray.INDArray labels, org.nd4j.linalg.api.ndarray.INDArray networkPredictions, org.nd4j.linalg.api.ndarray.INDArray maskArray)
long	getCountActualNegative(int outputNum) Get the actual negative count (accounting for any masking) for the specified output/column
long	getCountActualPositive(int outputNum) Get the actual positive count (accounting for any masking) for the specified output/column
PrecisionRecallCurve	getPrecisionRecallCurve(int outputNum) Get the Precision-Recall curve for the specified output
RocCurve	getRocCurve(int outputNum) Get the ROC curve for the specified output
void	merge(ROCBinary other)
int	numLabels() Returns the number of labels - (i.e., size of the prediction/labels arrays) - if known.
void	reset()
void	setLabelNames(List<String> labels) Set the label names, for printing via stats()
String	stats()
String	stats(int printPrecision)

Methods inherited from class org.deeplearning4j.eval.BaseEvaluation

equals, eval, evalTimeSeries, evalTimeSeries, fromJson, fromYaml, toJson, toString, toYaml

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_STATS_PRECISION

public static final int DEFAULT_STATS_PRECISION

See Also:

Constant Field Values

Constructor Detail

ROCBinary

public ROCBinary()

ROCBinary

public ROCBinary(int thresholdSteps)

Parameters:

thresholdSteps - Number of threshold steps to use for the ROC calculation. Set to 0 for exact ROC calculation

ROCBinary

public ROCBinary(int thresholdSteps,
                 boolean rocRemoveRedundantPts)

Parameters:

thresholdSteps - Number of threshold steps to use for the ROC calculation. If set to 0: use exact calculation

rocRemoveRedundantPts - Usually set to true. If true, remove any redundant points from ROC and P-R curves

Method Detail

reset

public void reset()

eval

public void eval(org.nd4j.linalg.api.ndarray.INDArray labels,
                 org.nd4j.linalg.api.ndarray.INDArray networkPredictions)

eval

public void eval(org.nd4j.linalg.api.ndarray.INDArray labels,
                 org.nd4j.linalg.api.ndarray.INDArray networkPredictions,
                 org.nd4j.linalg.api.ndarray.INDArray maskArray)

Specified by:

eval in interface IEvaluation<ROCBinary>

Overrides:

eval in class BaseEvaluation<ROCBinary>

merge

public void merge(ROCBinary other)

numLabels

public int numLabels()

Returns the number of labels - (i.e., size of the prediction/labels arrays) - if known. Returns -1 otherwise

getCountActualPositive

public long getCountActualPositive(int outputNum)

Get the actual positive count (accounting for any masking) for the specified output/column

Parameters:

outputNum - Index of the output (0 to numLabels()-1)

getCountActualNegative

public long getCountActualNegative(int outputNum)

Get the actual negative count (accounting for any masking) for the specified output/column

Parameters:

outputNum - Index of the output (0 to numLabels()-1)

getRocCurve

public RocCurve getRocCurve(int outputNum)

Get the ROC curve for the specified output

Parameters:

outputNum - Number of the output to get the ROC curve for

Returns:

ROC curve

getPrecisionRecallCurve

public PrecisionRecallCurve getPrecisionRecallCurve(int outputNum)

Get the Precision-Recall curve for the specified output

Parameters:

outputNum - Number of the output to get the P-R curve for

Returns:

Precision recall curve

calculateAverageAuc

public double calculateAverageAuc()

Macro-average AUC for all outcomes

Returns:

the (macro-)average AUC for all outcomes.

calculateAUC

public double calculateAUC(int outputNum)

Calculate the AUC - Area Under (ROC) Curve
Utilizes trapezoidal integration internally

Parameters:

outputNum - Output number to calculate AUC for

Returns:

AUC

setLabelNames

public void setLabelNames(List<String> labels)

Set the label names, for printing via stats()

stats

public String stats()

Returns:

stats

public String stats(int printPrecision)