/// <summary>Train a multinomial classifier off of the provided dataset.</summary>
/// <param name="dataset">The dataset to train the classifier off of.</param>
/// <returns>A classifier.</returns>
public static IClassifier <string, string> TrainMultinomialClassifier(GeneralDataset <string, string> dataset, int featureThreshold, double sigma)
{
// Set up the dataset and factory
log.Info("Applying feature threshold (" + featureThreshold + ")...");
dataset.ApplyFeatureCountThreshold(featureThreshold);
log.Info("Randomizing dataset...");
dataset.Randomize(42l);
log.Info("Creating factory...");
LinearClassifierFactory <string, string> factory = InitFactory(sigma);
// Train the final classifier
log.Info("BEGIN training");
LinearClassifier <string, string> classifier = factory.TrainClassifier(dataset);
log.Info("END training");
// Debug
KBPRelationExtractor.Accuracy trainAccuracy = new KBPRelationExtractor.Accuracy();
foreach (IDatum <string, string> datum in dataset)
{
string guess = classifier.ClassOf(datum);
trainAccuracy.Predict(Java.Util.Collections.Singleton(guess), Java.Util.Collections.Singleton(datum.Label()));
}
log.Info("Training accuracy:");
log.Info(trainAccuracy.ToString());
log.Info(string.Empty);
// Return the classifier
return(classifier);
}
/// <summary>Train on a list of ExtractionSentence containing labeled RelationMention objects</summary>
public virtual void Train(Annotation sentences)
{
// Train a single multi-class classifier
GeneralDataset <string, string> trainSet = CreateDataset(sentences);
TrainMulticlass(trainSet);
}
/// <exception cref="System.Exception"/>
public static void Main(string[] args)
{
Properties props;
if (args.Length > 0)
{
props = StringUtils.ArgsToProperties(args);
}
else
{
props = new Properties();
}
if (!props.Contains("dcoref.conll2011"))
{
log.Info("-dcoref.conll2011 [input_CoNLL_corpus]: was not specified");
return;
}
if (!props.Contains("singleton.predictor.output"))
{
log.Info("-singleton.predictor.output [output_model_file]: was not specified");
return;
}
GeneralDataset <string, string> data = Edu.Stanford.Nlp.Coref.Misc.SingletonPredictor.GenerateFeatureVectors(props);
LogisticClassifier <string, string> classifier = Edu.Stanford.Nlp.Coref.Misc.SingletonPredictor.Train(data);
Edu.Stanford.Nlp.Coref.Misc.SingletonPredictor.SaveToSerialized(classifier, GetPathSingletonPredictor(props));
}
public virtual void TrainMulticlass(GeneralDataset <string, string> trainSet)
{
if (Sharpen.Runtime.EqualsIgnoreCase(relationExtractorClassifierType, "linear"))
{
LinearClassifierFactory <string, string> lcFactory = new LinearClassifierFactory <string, string>(1e-4, false, sigma);
lcFactory.SetVerbose(false);
// use in-place SGD instead of QN. this is faster but much worse!
// lcFactory.useInPlaceStochasticGradientDescent(-1, -1, 1.0);
// use a hybrid minimizer: start with in-place SGD, continue with QN
// lcFactory.useHybridMinimizerWithInPlaceSGD(50, -1, sigma);
classifier = lcFactory.TrainClassifier(trainSet);
}
else
{
if (Sharpen.Runtime.EqualsIgnoreCase(relationExtractorClassifierType, "svm"))
{
SVMLightClassifierFactory <string, string> svmFactory = new SVMLightClassifierFactory <string, string>();
svmFactory.SetC(sigma);
classifier = svmFactory.TrainClassifier(trainSet);
}
else
{
throw new Exception("Invalid classifier type: " + relationExtractorClassifierType);
}
}
if (logger.IsLoggable(Level.Fine))
{
ReportWeights(classifier, null);
}
}
/// <summary>Train the singleton predictor using a logistic regression classifier.</summary>
/// <param name="pDataset">Dataset of features</param>
/// <returns>Singleton predictor</returns>
public static LogisticClassifier <string, string> Train(GeneralDataset <string, string> pDataset)
{
LogisticClassifierFactory <string, string> lcf = new LogisticClassifierFactory <string, string>();
LogisticClassifier <string, string> classifier = lcf.TrainClassifier(pDataset);
return(classifier);
}
public virtual double Score <F>(IClassifier <L, F> classifier, GeneralDataset <L, F> data)
{
IList <L> guesses = new List <L>();
IList <L> labels = new List <L>();
for (int i = 0; i < data.Size(); i++)
{
IDatum <L, F> d = data.GetRVFDatum(i);
L guess = classifier.ClassOf(d);
guesses.Add(guess);
}
int[] labelsArr = data.GetLabelsArray();
labelIndex = data.labelIndex;
for (int i_1 = 0; i_1 < data.Size(); i_1++)
{
labels.Add(labelIndex.Get(labelsArr[i_1]));
}
labelIndex = new HashIndex <L>();
labelIndex.AddAll(data.LabelIndex().ObjectsList());
labelIndex.AddAll(classifier.Labels());
int numClasses = labelIndex.Size();
tpCount = new int[numClasses];
fpCount = new int[numClasses];
fnCount = new int[numClasses];
negIndex = labelIndex.IndexOf(negLabel);
for (int i_2 = 0; i_2 < guesses.Count; ++i_2)
{
L guess = guesses[i_2];
int guessIndex = labelIndex.IndexOf(guess);
L label = labels[i_2];
int trueIndex = labelIndex.IndexOf(label);
if (guessIndex == trueIndex)
{
if (guessIndex != negIndex)
{
tpCount[guessIndex]++;
}
}
else
{
if (guessIndex != negIndex)
{
fpCount[guessIndex]++;
}
if (trueIndex != negIndex)
{
fnCount[trueIndex]++;
}
}
}
return(GetFMeasure());
}
public override double Score <F>(IClassifier <L, F> classifier, GeneralDataset <L, F> data)
{
labelIndex = new HashIndex <L>();
labelIndex.AddAll(classifier.Labels());
labelIndex.AddAll(data.labelIndex.ObjectsList());
ClearCounts();
int[] labelsArr = data.GetLabelsArray();
for (int i = 0; i < data.Size(); i++)
{
IDatum <L, F> d = data.GetRVFDatum(i);
L guess = classifier.ClassOf(d);
AddGuess(guess, labelIndex.Get(labelsArr[i]));
}
FinalizeCounts();
return(GetFMeasure());
}
public virtual void InitMC <F>(IProbabilisticClassifier <L, F> classifier, GeneralDataset <L, F> data)
{
//if (!(gData instanceof Dataset)) {
// throw new UnsupportedOperationException("Can only handle Datasets, not "+gData.getClass().getName());
//}
//
//Dataset data = (Dataset)gData;
IPriorityQueue <Pair <int, Pair <double, bool> > > q = new BinaryHeapPriorityQueue <Pair <int, Pair <double, bool> > >();
total = 0;
correct = 0;
logLikelihood = 0.0;
for (int i = 0; i < data.Size(); i++)
{
IDatum <L, F> d = data.GetRVFDatum(i);
ICounter <L> scores = classifier.LogProbabilityOf(d);
L guess = Counters.Argmax(scores);
L correctLab = d.Label();
double guessScore = scores.GetCount(guess);
double correctScore = scores.GetCount(correctLab);
int guessInd = data.LabelIndex().IndexOf(guess);
int correctInd = data.LabelIndex().IndexOf(correctLab);
total++;
if (guessInd == correctInd)
{
correct++;
}
logLikelihood += correctScore;
q.Add(new Pair <int, Pair <double, bool> >(int.Parse(i), new Pair <double, bool>(guessScore, bool.ValueOf(guessInd == correctInd))), -guessScore);
}
accuracy = (double)correct / (double)total;
IList <Pair <int, Pair <double, bool> > > sorted = q.ToSortedList();
scores = new double[sorted.Count];
isCorrect = new bool[sorted.Count];
for (int i_1 = 0; i_1 < sorted.Count; i_1++)
{
Pair <double, bool> next = sorted[i_1].Second();
scores[i_1] = next.First();
isCorrect[i_1] = next.Second();
}
}
public virtual double Score <F>(IProbabilisticClassifier <L, F> classifier, GeneralDataset <L, F> data)
{
List <Pair <double, int> > dataScores = new List <Pair <double, int> >();
for (int i = 0; i < data.Size(); i++)
{
IDatum <L, F> d = data.GetRVFDatum(i);
ICounter <L> scores = classifier.LogProbabilityOf(d);
int labelD = d.Label().Equals(posLabel) ? 1 : 0;
dataScores.Add(new Pair <double, int>(Math.Exp(scores.GetCount(posLabel)), labelD));
}
PRCurve prc = new PRCurve(dataScores);
confWeightedAccuracy = prc.Cwa();
accuracy = prc.Accuracy();
optAccuracy = prc.OptimalAccuracy();
optConfWeightedAccuracy = prc.OptimalCwa();
logLikelihood = prc.LogLikelihood();
accrecall = prc.CwaArray();
optaccrecall = prc.OptimalCwaArray();
return(accuracy);
}
public MultiClassChunkEvalStats(IClassifier <string, F> classifier, GeneralDataset <string, F> data, string negLabel)
: base(classifier, data, negLabel)
{
chunker = new LabeledChunkIdentifier();
chunker.SetNegLabel(negLabel);
}
public BiasedLogConditionalObjectiveFunction(GeneralDataset dataset, double[][] confusionMatrix)
: this(dataset, confusionMatrix, new LogPrior(LogPrior.LogPriorType.QUADRATIC))
{
}
public MultiClassStringLabelStats(IClassifier <string, F> classifier, GeneralDataset <string, F> data, string negLabel)
: base(classifier, data, negLabel)
{
stringConverter = new MultiClassPrecisionRecallExtendedStats.StringStringConverter();
}
public BiasedLogConditionalObjectiveFunction(GeneralDataset dataset, double[][] confusionMatrix, LogPrior prior)
: this(dataset.NumFeatures(), dataset.NumClasses(), dataset.GetDataArray(), dataset.GetLabelsArray(), confusionMatrix, prior)
{
}
public MultiClassAccuracyStats(IProbabilisticClassifier <L, F> classifier, GeneralDataset <L, F> data, string file, int scoreType)
{
saveFile = file;
this.scoreType = scoreType;
InitMC(classifier, data);
}
/// <summary>A helper function for dumping the accuracy of the trained classifier.</summary>
/// <param name="classifier">The classifier to evaluate.</param>
/// <param name="dataset">The dataset to evaluate the classifier on.</param>
public static void DumpAccuracy(IClassifier <ClauseSplitter.ClauseClassifierLabel, string> classifier, GeneralDataset <ClauseSplitter.ClauseClassifierLabel, string> dataset)
{
DecimalFormat df = new DecimalFormat("0.00%");
Redwood.Log("size: " + dataset.Size());
Redwood.Log("split count: " + StreamSupport.Stream(dataset.Spliterator(), false).Filter(null).Collect(Collectors.ToList()).Count);
Redwood.Log("interm count: " + StreamSupport.Stream(dataset.Spliterator(), false).Filter(null).Collect(Collectors.ToList()).Count);
Pair <double, double> pr = classifier.EvaluatePrecisionAndRecall(dataset, ClauseSplitter.ClauseClassifierLabel.ClauseSplit);
Redwood.Log("p (split): " + df.Format(pr.first));
Redwood.Log("r (split): " + df.Format(pr.second));
Redwood.Log("f1 (split): " + df.Format(2 * pr.first * pr.second / (pr.first + pr.second)));
pr = classifier.EvaluatePrecisionAndRecall(dataset, ClauseSplitter.ClauseClassifierLabel.ClauseInterm);
Redwood.Log("p (interm): " + df.Format(pr.first));
Redwood.Log("r (interm): " + df.Format(pr.second));
Redwood.Log("f1 (interm): " + df.Format(2 * pr.first * pr.second / (pr.first + pr.second)));
}
public virtual double Score <F>(IProbabilisticClassifier <L, F> classifier, GeneralDataset <L, F> data)
{
return(Score((IClassifier <L, F>)classifier, data));
}
public MultiClassPrecisionRecallStats(IClassifier <L, F> classifier, GeneralDataset <L, F> data, L negLabel)
{
this.negLabel = negLabel;
Score(classifier, data);
}
public FeaturesData(IDictionary <int, Pair <int, int> > mapQuoteToDataRange, IDictionary <int, Sieve.MentionData> mapDatumToMention, GeneralDataset <string, string> dataset)
{
this.mapQuoteToDataRange = mapQuoteToDataRange;
this.mapDatumToMention = mapDatumToMention;
this.dataset = dataset;
}
public AccuracyStats(IProbabilisticClassifier <L, F> classifier, GeneralDataset <L, F> data, L posLabel)
{
// = null;
this.posLabel = posLabel;
Score(classifier, data);
}
public MultiClassAccuracyStats(IProbabilisticClassifier <L, F> classifier, GeneralDataset <L, F> data, string file)
: this(classifier, data, file, UseAccuracy)
{
}
public MultiClassPrecisionRecallExtendedStats(IClassifier <L, F> classifier, GeneralDataset <L, F> data, L negLabel)
: base(classifier, data, negLabel)
{
}
public virtual double Score <F>(IProbabilisticClassifier <L, F> classifier, GeneralDataset <L, F> data)
{
InitMC(classifier, data);
return(Score());
}
public ExtractQuotesClassifier(GeneralDataset <string, string> trainingSet)
{
LinearClassifierFactory <string, string> lcf = new LinearClassifierFactory <string, string>();
quoteToMentionClassifier = lcf.TrainClassifier(trainingSet);
}