com.emphysic.myriad.core.data.roi

Class MLROIConfFinder

java.lang.Object

com.emphysic.myriad.core.data.roi.MLROIConfFinder

All Implemented Interfaces:

MLROIFinder, ROIFinder, ROIProbability, ObjectMap, com.esotericsoftware.kryo.KryoSerializable

Direct Known Subclasses:

AdaptiveSGDROIFinder, GradMachineROIFinder, PassiveAggressiveROIFinder

public abstract class MLROIConfFinder
extends java.lang.Object
implements MLROIFinder, ROIProbability

A Region of Interest (ROI) finder based on machine learning that provides both probabilities of its classifications and the ability to define a confidence threshold. Created by ccoughlin on 9/21/2016.

Field Summary

Fields

Modifier and Type	Field and Description
protected double	confThr A confidence threshold between 0 and 1 to be met for labelling a sample as containing ROI.

Constructor Summary

Constructors

Constructor and Description
MLROIConfFinder()

Method Summary

Modifier and Type	Method and Description
double	getConfidenceThreshold() Returns the confidence threshold for labelling ROI.
double	predict(Dataset data) Predicts the label for a sample.
double	predict(double[] data) Predicts the label for a sample.
void	setConfidenceThreshold(double confThr) Sets the confidence threshold for labelling ROI samples - the model must be (confThr * 100)% confident a sample contains ROI for it to be labelled as such.
static byte[]	writableWriteToBytes(org.apache.hadoop.io.Writable writable) Convenience method for writing a Mahout Writable to a byte array

Methods inherited from class java.lang.Object

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

Methods inherited from interface com.emphysic.myriad.core.data.roi.MLROIFinder

negativeClass, positiveClass, train

Methods inherited from interface com.emphysic.myriad.core.data.roi.ROIFinder

fromFile, isROI, isROI

Methods inherited from interface com.emphysic.myriad.core.data.util.ObjectMap

getObjectMap, getSerializationVersion, getVersion, init, initCurrentVersion, initPreviousVersion, initUnknownVersion, load, read, save, write

Methods inherited from interface com.emphysic.myriad.core.data.roi.ROIProbability

classify, predict_proba, predict_proba

Field Detail

confThr

protected double confThr

A confidence threshold between 0 and 1 to be met for labelling a sample as containing ROI. The higher the threshold the fewer ROI will be reported. Defaults to 0 i.e. no confidence thresholding is performed.

Constructor Detail

MLROIConfFinder

public MLROIConfFinder()

Method Detail

getConfidenceThreshold

public double getConfidenceThreshold()

Returns the confidence threshold for labelling ROI.

Returns:

confidence threshold between 0 and 1

setConfidenceThreshold

public void setConfidenceThreshold(double confThr)
                            throws java.lang.IllegalArgumentException

Sets the confidence threshold for labelling ROI samples - the model must be (confThr * 100)% confident a sample contains ROI for it to be labelled as such.

Parameters:

confThr - new confidence threshold

Throws:

java.lang.IllegalArgumentException - if the confidence threshold is not in the range 0-1 inclusive.

predict

public double predict(double[] data)

Predicts the label for a sample. Prediction is the category index with the maximum value.

Specified by:

predict in interface MLROIFinder

Parameters:

data - sample to predict

Returns:

integer index with the maximum value

predict

public double predict(Dataset data)

Predicts the label for a sample. Prediction is the category index with the maximum value.

Specified by:

predict in interface MLROIFinder

Parameters:

data - sample to predict

Returns:

integer index with the maximum value

writableWriteToBytes

public static byte[] writableWriteToBytes(org.apache.hadoop.io.Writable writable)

Convenience method for writing a Mahout Writable to a byte array

Parameters:

writable - Mahout object that implements the Writable interface

Returns:

byte array