/*
* EvaluatePeriodicHeldOutTest.java
* Copyright (C) 2007 University of Waikato, Hamilton, New Zealand
* @author Richard Kirkby (rkirkby@cs.waikato.ac.nz)
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
package tr.gov.ulakbim.jDenetX.tasks;
import tr.gov.ulakbim.jDenetX.classifiers.Classifier;
import tr.gov.ulakbim.jDenetX.core.Measurement;
import tr.gov.ulakbim.jDenetX.core.ObjectRepository;
import tr.gov.ulakbim.jDenetX.core.StringUtils;
import tr.gov.ulakbim.jDenetX.core.TimingUtils;
import tr.gov.ulakbim.jDenetX.evaluation.ClassificationPerformanceEvaluator;
import tr.gov.ulakbim.jDenetX.evaluation.LearningCurve;
import tr.gov.ulakbim.jDenetX.evaluation.LearningEvaluation;
import tr.gov.ulakbim.jDenetX.options.ClassOption;
import tr.gov.ulakbim.jDenetX.options.FileOption;
import tr.gov.ulakbim.jDenetX.options.FlagOption;
import tr.gov.ulakbim.jDenetX.options.IntOption;
import tr.gov.ulakbim.jDenetX.streams.CachedInstancesStream;
import tr.gov.ulakbim.jDenetX.streams.InstanceStream;
import weka.core.Instance;
import weka.core.Instances;
import java.io.File;
import java.io.FileOutputStream;
import java.io.PrintStream;
import java.util.ArrayList;
import java.util.List;
public class EvaluatePeriodicHeldOutTest extends MainTask {
@Override
public String getPurposeString() {
return "Evaluates a classifier on a stream by periodically testing on a heldout set.";
}
private static final long serialVersionUID = 1L;
public ClassOption learnerOption = new ClassOption("learner", 'l',
"Classifier to train.", Classifier.class, "HoeffdingTree");
public ClassOption streamOption = new ClassOption("stream", 's',
"Stream to learn from.", InstanceStream.class,
"generators.RandomTreeGenerator");
public ClassOption evaluatorOption = new ClassOption("evaluator", 'e',
"Classification performance evaluation method.",
ClassificationPerformanceEvaluator.class,
"BasicClassificationPerformanceEvaluator");
public IntOption testSizeOption = new IntOption("testSize", 'n',
"Number of testing examples.", 1000000, 0, Integer.MAX_VALUE);
public IntOption trainSizeOption = new IntOption("trainSize", 'i',
"Number of training examples, <1 = unlimited.", 0, 0,
Integer.MAX_VALUE);
public IntOption trainTimeOption = new IntOption("trainTime", 't',
"Number of training seconds.", 10 * 60 * 60, 0, Integer.MAX_VALUE);
public IntOption sampleFrequencyOption = new IntOption(
"sampleFrequency",
'f',
"Number of training examples between samples of learning performance.",
100000, 0, Integer.MAX_VALUE);
public FileOption dumpFileOption = new FileOption("dumpFile", 'd',
"File to append intermediate csv results to.", null, "csv", true);
public FlagOption cacheTestOption = new FlagOption("cacheTest", 'c',
"Cache test instances in memory.");
@Override
protected Object doMainTask(TaskMonitor monitor, ObjectRepository repository) {
Classifier learner = (Classifier) getPreparedClassOption(this.learnerOption);
InstanceStream stream = (InstanceStream) getPreparedClassOption(this.streamOption);
ClassificationPerformanceEvaluator evaluator = (ClassificationPerformanceEvaluator) getPreparedClassOption(this.evaluatorOption);
learner.setModelContext(stream.getHeader());
long instancesProcessed = 0;
LearningCurve learningCurve = new LearningCurve("evaluation instances");
File dumpFile = this.dumpFileOption.getFile();
PrintStream immediateResultStream = null;
if (dumpFile != null) {
try {
if (dumpFile.exists()) {
immediateResultStream = new PrintStream(
new FileOutputStream(dumpFile, true), true);
} else {
immediateResultStream = new PrintStream(
new FileOutputStream(dumpFile), true);
}
} catch (Exception ex) {
throw new RuntimeException(
"Unable to open immediate result file: " + dumpFile, ex);
}
}
boolean firstDump = true;
InstanceStream testStream = null;
int testSize = this.testSizeOption.getValue();
if (this.cacheTestOption.isSet()) {
monitor.setCurrentActivity("Caching test examples...", -1.0);
Instances testInstances = new Instances(stream.getHeader(),
this.testSizeOption.getValue());
while (testInstances.numInstances() < testSize) {
testInstances.add(stream.nextInstance());
if (testInstances.numInstances()
% INSTANCES_BETWEEN_MONITOR_UPDATES == 0) {
if (monitor.taskShouldAbort()) {
return null;
}
monitor
.setCurrentActivityFractionComplete((double) testInstances
.numInstances()
/ (double) (this.testSizeOption.getValue()));
}
}
testStream = new CachedInstancesStream(testInstances);
} else {
testStream = (InstanceStream) stream.copy();
monitor.setCurrentActivity("Skipping test examples...", -1.0);
for (int i = 0; i < testSize; i++) {
stream.nextInstance();
}
}
instancesProcessed = 0;
TimingUtils.enablePreciseTiming();
double totalTrainTime = 0.0;
while ((this.trainSizeOption.getValue() < 1)
|| (instancesProcessed < this.trainSizeOption.getValue())) {
monitor.setCurrentActivityDescription("Training...");
long instancesTarget = instancesProcessed
+ this.sampleFrequencyOption.getValue();
long trainStartTime = TimingUtils.getNanoCPUTimeOfCurrentThread();
while (instancesProcessed < instancesTarget) {
learner.trainOnInstance(stream.nextInstance());
instancesProcessed++;
if (instancesProcessed % INSTANCES_BETWEEN_MONITOR_UPDATES == 0) {
if (monitor.taskShouldAbort()) {
return null;
}
monitor
.setCurrentActivityFractionComplete((double) (instancesProcessed)
/ (double) (this.trainSizeOption.getValue()));
}
}
double lastTrainTime = TimingUtils.nanoTimeToSeconds(TimingUtils
.getNanoCPUTimeOfCurrentThread()
- trainStartTime);
totalTrainTime += lastTrainTime;
if (totalTrainTime > this.trainTimeOption.getValue()) {
break;
}
testStream.restart();
evaluator.reset();
long testInstancesProcessed = 0;
monitor.setCurrentActivityDescription("Testing (after "
+ StringUtils
.doubleToString(
((double) (instancesProcessed)
/ (double) (this.trainSizeOption
.getValue()) * 100.0), 2)
+ "% training)...");
long testStartTime = TimingUtils.getNanoCPUTimeOfCurrentThread();
for (int i = 0; i < testSize; i++) {
Instance testInst = (Instance) testStream.nextInstance().copy();
int trueClass = (int) testInst.classValue();
testInst.setClassMissing();
double[] prediction = learner.getVotesForInstance(testInst);
evaluator.addClassificationAttempt(trueClass, prediction,
testInst.weight());
testInstancesProcessed++;
if (testInstancesProcessed % INSTANCES_BETWEEN_MONITOR_UPDATES == 0) {
if (monitor.taskShouldAbort()) {
return null;
}
monitor
.setCurrentActivityFractionComplete((double) testInstancesProcessed
/ (double) (testSize));
}
}
double testTime = TimingUtils.nanoTimeToSeconds(TimingUtils
.getNanoCPUTimeOfCurrentThread()
- testStartTime);
List<Measurement> measurements = new ArrayList<Measurement>();
measurements.add(new Measurement("evaluation instances",
instancesProcessed));
measurements
.add(new Measurement("total train time", totalTrainTime));
measurements.add(new Measurement("total train speed",
instancesProcessed / totalTrainTime));
measurements.add(new Measurement("last train time", lastTrainTime));
measurements.add(new Measurement("last train speed",
this.sampleFrequencyOption.getValue() / lastTrainTime));
measurements.add(new Measurement("test time", testTime));
measurements.add(new Measurement("test speed", this.testSizeOption
.getValue()
/ testTime));
Measurement[] performanceMeasurements = evaluator
.getPerformanceMeasurements();
for (Measurement measurement : performanceMeasurements) {
measurements.add(measurement);
}
Measurement[] modelMeasurements = learner.getModelMeasurements();
for (Measurement measurement : modelMeasurements) {
measurements.add(measurement);
}
learningCurve.insertEntry(new LearningEvaluation(measurements
.toArray(new Measurement[measurements.size()])));
if (immediateResultStream != null) {
if (firstDump) {
immediateResultStream.println(learningCurve
.headerToString());
firstDump = false;
}
immediateResultStream.println(learningCurve
.entryToString(learningCurve.numEntries() - 1));
immediateResultStream.flush();
}
if (monitor.resultPreviewRequested()) {
monitor.setLatestResultPreview(learningCurve.copy());
}
// if (learner instanceof HoeffdingTree
// || learner instanceof HoeffdingOptionTree) {
// int numActiveNodes = (int) Measurement.getMeasurementNamed(
// "active learning leaves",
// modelMeasurements).getValue();
// // exit if tree frozen
// if (numActiveNodes < 1) {
// break;
// }
// int numNodes = (int) Measurement.getMeasurementNamed(
// "tree size (nodes)", modelMeasurements)
// .getValue();
// if (numNodes == lastNumNodes) {
// noGrowthCount++;
// } else {
// noGrowthCount = 0;
// }
// lastNumNodes = numNodes;
// } else if (learner instanceof OzaBoost || learner instanceof
// OzaBag) {
// double numActiveNodes = Measurement.getMeasurementNamed(
// "[avg] active learning leaves",
// modelMeasurements).getValue();
// // exit if all trees frozen
// if (numActiveNodes == 0.0) {
// break;
// }
// int numNodes = (int) (Measurement.getMeasurementNamed(
// "[avg] tree size (nodes)",
// learner.getModelMeasurements()).getValue() * Measurement
// .getMeasurementNamed("ensemble size",
// modelMeasurements).getValue());
// if (numNodes == lastNumNodes) {
// noGrowthCount++;
// } else {
// noGrowthCount = 0;
// }
// lastNumNodes = numNodes;
// }
}
if (immediateResultStream != null) {
immediateResultStream.close();
}
return learningCurve;
}
public Class<?> getTaskResultType() {
return LearningCurve.class;
}
}