/**
* resets the classifier, i.e., the best model, last known restart and iteration, etc.
*
* @see #m_BestModel
* @see #m_LastIteration
* @see #m_LastRestart
*/
@Override
public void reset() {
super.reset();
m_BestModel = null;
m_LastIteration = -1;
m_LastRestart = -1;
m_LastRMSTrain = 1.0;
m_LastRMSTest = 1.0;
m_LastRMSTestOriginal = 1.0;
m_LastRMS = 1.0;
m_RMSTrain = 1.0;
m_RMSTest = 1.0;
m_RMSTestOriginal = 1.0;
m_RMS = 1.0;
m_AccTrain = 0.0;
m_AccTestOriginal = 0.0;
m_LastAccTrain = 0.0;
// for logging
m_LogEntries.clear();
m_LogEntries.addFilename("RMS", createFilename("-rms.csv"));
m_LogEntries.addFilename("RMSTrain", createFilename("-rms_train.csv"));
m_LogEntries.addFilename("RMSTest", createFilename("-rms_test.csv"));
m_LogEntries.addFilename("RMSTestOriginal", createFilename("-rms_test-original.csv"));
m_LogEntries.addFilename("AccTrain", createFilename("-acc_train.csv"));
m_LogEntries.addFilename("AccTestOriginal", createFilename("-acc_test-original.csv"));
m_LogEntries.addFilename("FlippedLabels", createFilename("-flipped.csv"));
}
/**
* builds the necessary CollectiveInstances from the given Instances
*
* @throws Exception if anything goes wrong
*/
@Override
protected void generateSets() throws Exception {
int i;
super.generateSets();
m_CollectiveInstances = new CollectiveInstances();
m_CollectiveInstances.setSeed(getSeed());
m_CollectiveInstances.setFlipper(m_Flipper);
m_TrainsetNew = new Instances(m_Trainset);
for (i = 0; i < m_Testset.numInstances(); i++) m_TrainsetNew.add(m_Testset.instance(i));
m_FlipHistory = new FlipHistory(m_TrainsetNew);
}
/**
* Parses a given list of options.
*
* <p>
* <!-- options-start -->
* Valid options are:
*
* <p>
*
* <pre> -I <num>
* Number of iterations.
* (default 10)</pre>
*
* <pre> -R <num>
* Number of restarts.
* (default 10)</pre>
*
* <pre> -log
* Creates logs in the tmp directory for all kinds of internal data.
* Use only for debugging purposes!
* </pre>
*
* <pre> -U
* Updates also the labels of the training set.
* </pre>
*
* <pre> -eval <num>
* The type of evaluation to use (0 = Randomwalk/Last model used for
* prediction, 1=Randomwalk/Best model used for prediction,
* 2=Hillclimbing).
* </pre>
*
* <pre> -compare <num>
* The type of comparisong used for comparing models.
* (0=overall RMS, 1=RMS on train set, 2=RMS on test set,
* 3=Accuracy on train set)
* </pre>
*
* <pre> -flipper "<classname [parameters]>"
* The flipping algorithm (and optional parameters) to use for
* flipping labels.
* </pre>
*
* <pre> -folds <folds>
* The number of folds for splitting the training set into
* train and test set. The first fold is always the training
* set. With '-V' you can invert this, i.e., instead of 20/80
* for 5 folds you'll get 80/20.
* (default 5)</pre>
*
* <pre> -V
* Inverts the fold selection, i.e., instead of using the first
* fold for the training set it is used for test set and the
* remaining folds for training.</pre>
*
* <pre> -verbose
* Whether to print some more information during building the
* classifier.
* (default is off)</pre>
*
* <pre> -verbose
* Whether to print some more information during building the
* classifier.
* (default is off)</pre>
*
* <pre> -S <num>
* Random number seed.
* (default 1)</pre>
*
* <pre> -D
* If set, classifier is run in debug mode and
* may output additional info to the console</pre>
*
* <pre> -W
* Full name of base classifier.
* (default: weka.classifiers.trees.J48)</pre>
*
* <pre>
* Options specific to classifier weka.classifiers.trees.J48:
* </pre>
*
* <pre> -U
* Use unpruned tree.</pre>
*
* <pre> -C <pruning confidence>
* Set confidence threshold for pruning.
* (default 0.25)</pre>
*
* <pre> -M <minimum number of instances>
* Set minimum number of instances per leaf.
* (default 2)</pre>
*
* <pre> -R
* Use reduced error pruning.</pre>
*
* <pre> -N <number of folds>
* Set number of folds for reduced error
* pruning. One fold is used as pruning set.
* (default 3)</pre>
*
* <pre> -B
* Use binary splits only.</pre>
*
* <pre> -S
* Don't perform subtree raising.</pre>
*
* <pre> -L
* Do not clean up after the tree has been built.</pre>
*
* <pre> -A
* Laplace smoothing for predicted probabilities.</pre>
*
* <pre> -Q <seed>
* Seed for random data shuffling (default 1).</pre>
*
* <!-- options-end -->
* Options after -- are passed to the designated classifier.
*
* <p>
*
* @param options the list of options as an array of strings
* @throws Exception if an option is not supported
*/
@Override
public void setOptions(String[] options) throws Exception {
String tmpStr;
String[] tmpOptions;
tmpStr = Utils.getOption('I', options);
if (tmpStr.length() != 0) setNumIterations(Integer.parseInt(tmpStr));
else setNumIterations(10);
tmpStr = Utils.getOption('R', options);
if (tmpStr.length() != 0) setNumRestarts(Integer.parseInt(tmpStr));
else setNumRestarts(10);
setLog(Utils.getFlag("log", options));
setUpdateTraining(Utils.getFlag('U', options));
tmpStr = Utils.getOption("eval", options);
if (tmpStr.length() != 0)
setEvaluationType(new SelectedTag(Integer.parseInt(tmpStr), CollectiveInstances.EVAL_TAGS));
else
setEvaluationType(
new SelectedTag(CollectiveInstances.EVAL_RANDOMWALK_LAST, CollectiveInstances.EVAL_TAGS));
tmpStr = Utils.getOption("compare", options);
if (tmpStr.length() != 0)
setComparisonType(
new SelectedTag(Integer.parseInt(tmpStr), CollectiveInstances.COMPARE_TAGS));
else
setComparisonType(
new SelectedTag(CollectiveInstances.COMPARE_RMS, CollectiveInstances.COMPARE_TAGS));
tmpStr = Utils.getOption("flipper", options);
if (tmpStr.length() != 0) {
tmpOptions = Utils.splitOptions(tmpStr);
tmpStr = tmpOptions[0];
tmpOptions[0] = "";
setFlipper(Flipper.forName(tmpStr, tmpOptions));
} else {
setFlipper(new TriangleFlipper());
}
super.setOptions(options);
}
/** performs initialization of members */
@Override
protected void initializeMembers() {
super.initializeMembers();
m_NumIterations = 10;
m_NumRestarts = 10;
m_LastRestart = -1;
m_LastIteration = -1;
m_RMSTrain = 1.0;
m_RMSTest = 1.0;
m_RMSTestOriginal = 1.0;
m_RMS = 1.0;
m_AccTrain = 0.0;
m_AccTestOriginal = 0.0;
m_LastRMSTrain = 1.0;
m_LastRMSTest = 1.0;
m_LastRMSTestOriginal = 1.0;
m_LastRMS = 1.0;
m_LastAccTrain = 0.0;
m_BestModel = null;
m_CollectiveInstances = null;
m_UpdateTraining = false;
m_EvaluationType = CollectiveInstances.EVAL_RANDOMWALK_BEST;
m_ComparisonType = CollectiveInstances.COMPARE_RMS_TRAIN;
m_Log = false;
m_TrainsetNew = null;
m_FlipHistory = null;
m_Classifier = new weka.classifiers.trees.J48();
m_Flipper = new TriangleFlipper();
m_LogEntries = new CollectiveLog();
m_AdditionalMeasures.add("measureLastRestart");
m_AdditionalMeasures.add("measureLastIteration");
m_AdditionalMeasures.add("measureLastRMSTrain");
m_AdditionalMeasures.add("measureLastRMSTest");
m_AdditionalMeasures.add("measureLastRMSTestOriginal");
m_AdditionalMeasures.add("measureLastRMS");
m_AdditionalMeasures.add("measureLastAccTrain");
}