/// <summary>
/// Measure quality of the training parameters.
/// </summary>
/// <param name="parameters">The parameters.</param>
/// <returns>The evaluated result.</returns>
public double Evaluate(double[] parameters)
{
var contexts = indexer.GetContexts();
var values = indexer.Values;
var nEventsSeen = indexer.GetNumTimesEventsSeen();
var outcomeList = indexer.GetOutcomeList();
var nOutcomes = outcomeList.Length;
var nPredLabels = indexer.GetPredLabels().Length;
var nCorrect = 0;
var nTotalEvents = 0;
for (var ei = 0; ei < contexts.Length; ei++)
{
var context = contexts[ei];
var value = values == null ? null : values[ei];
var probs = new double[nOutcomes];
QNModel.Eval(context, value, probs, nOutcomes, nPredLabels, parameters);
var outcome = ArrayMath.MaxId(probs);
if (outcome == outcomeList[ei])
{
nCorrect += nEventsSeen[ei];
}
nTotalEvents += nEventsSeen[ei];
}
return(nTotalEvents == 0 ? 0 : (double)nCorrect / nTotalEvents);
}
/// <summary>
/// Execute the training operation.
/// </summary>
/// <param name="indexer">The data indexer.</param>
/// <returns>The trained <see cref="IMaxentModel"/> model.</returns>
protected override IMaxentModel DoTrain(IDataIndexer indexer)
{
Display("Incorporating indexed data for training...");
indexer.Execute();
contexts = indexer.GetContexts();
values = indexer.Values;
numTimesEventsSeen = indexer.GetNumTimesEventsSeen();
numEvents = indexer.GetNumEvents();
numUniqueEvents = contexts.Length;
outcomeLabels = indexer.GetOutcomeLabels();
outcomeList = indexer.GetOutcomeList();
predLabels = indexer.GetPredLabels();
numPreds = predLabels.Length;
numOutcomes = outcomeLabels.Length;
Display("done.");
Display("\tNumber of Event Tokens: " + numUniqueEvents);
Display("\t Number of Outcomes: " + numOutcomes);
Display("\t Number of Predicates: " + numPreds);
Display("Computing model parameters...");
// ReSharper disable once CoVariantArrayConversion - we read the parameters ;)
Context[] finalParameters = FindParameters();
Display("...done.\n");
return(new NaiveBayesModel(finalParameters, predLabels, outcomeLabels));
}
/// <summary>
/// Initializes a new instance of the <see cref="NegLogLikelihood"/> class.
/// </summary>
/// <param name="indexer">The data indexer.</param>
public NegLogLikelihood(IDataIndexer indexer)
{
if (indexer == null)
{
throw new ArgumentNullException(nameof(indexer));
}
if (!indexer.Completed)
{
indexer.Execute();
}
values = indexer is OnePassRealValueDataIndexer ? indexer.Values : null;
contexts = indexer.GetContexts();
outcomeList = indexer.GetOutcomeList();
numTimesEventsSeen = indexer.GetNumTimesEventsSeen();
numOutcomes = indexer.GetOutcomeLabels().Length;
numFeatures = indexer.GetPredLabels().Length;
numContexts = contexts.Length;
dimension = numOutcomes * numFeatures;
expectation = new double[numOutcomes];
tempSums = new double[numOutcomes];
gradient = new double[dimension];
}
/// <summary>
/// Train a model using the Perceptron algorithm.
/// </summary>
/// <param name="iterations">The number of Perceptron iterations to perform.</param>
/// <param name="indexer">The object which will be used for event compilation.</param>
/// <param name="cutoff">The number of times a predicate must occur to be used in a model.</param>
/// <param name="useAverage"></param>
/// <returns>The newly trained model, which can be used immediately or saved to disk using a <see cref="IO.PerceptronModelWriter"/> object.</returns>
public AbstractModel TrainModel(int iterations, IDataIndexer indexer, int cutoff, bool useAverage)
{
Display("Incorporating indexed data for training...");
info.Append("Trained using Perceptron algorithm.\n\n");
// Executes the data indexer
indexer.Execute();
contexts = indexer.GetContexts();
values = indexer.Values;
numTimesEventsSeen = indexer.GetNumTimesEventsSeen();
numEvents = indexer.GetNumEvents();
numUniqueEvents = contexts.Length;
outcomeLabels = indexer.GetOutcomeLabels();
outcomeList = indexer.GetOutcomeList();
predLabels = indexer.GetPredLabels();
numPreds = predLabels.Length;
numOutcomes = outcomeLabels.Length;
Display("\ndone.\n");
info.Append("Number of Event Tokens: {0}\n" +
" Number of Outcomes: {1}\n" +
" Number of Predicates: {2}\n", numEvents, numOutcomes, numPreds);
Display("\tNumber of Event Tokens: " + numUniqueEvents);
Display("\t Number of Outcomes: " + numOutcomes);
Display("\t Number of Predicates: " + numPreds);
Display("Computing model parameters.");
var finalParameters = FindParameters(iterations, useAverage);
Display("\ndone.\n");
// ReSharper disable once CoVariantArrayConversion
return(new PerceptronModel(finalParameters, predLabels, outcomeLabels)
{
info = info
});
}
private bool CompareDoubleArray(double[] expected, double[] actual, IDataIndexer indexer, double tolerance)
{
var alignedActual = AlignDoubleArrayForTestData(actual, indexer.GetPredLabels(), indexer.GetOutcomeLabels());
if (expected.Length != alignedActual.Length)
{
return(false);
}
for (var i = 0; i < alignedActual.Length; i++)
{
if (Math.Abs(alignedActual[i] - expected[i]) > tolerance)
{
return(false);
}
}
return(true);
}
/// <summary>
/// Train a model using the GIS algorithm.
/// </summary>
/// <param name="iterations">The number of GIS iterations to perform.</param>
/// <param name="di">The data indexer used to compress events in memory.</param>
/// <param name="modelPrior">The prior distribution used to train this model.</param>
/// <param name="modelCutoff">The number of times a feature must occur to be included.</param>
/// <param name="threads">The number of threads used to train this model.</param>
/// <returns>The newly trained model, which can be used immediately or saved to disk using an <see cref="GISModelWriter"/> object.</returns>
public GISModel TrainModel(int iterations, IDataIndexer di, IPrior modelPrior, int modelCutoff, int threads) {
if (threads <= 0)
throw new ArgumentOutOfRangeException("threads", threads, @"Threads must be at least one or greater.");
modelExpects = new MutableContext[threads][];
info.Append("Trained using GIS algorithm.\n\n");
// Executes the data indexer
di.Execute();
// Incorporate all of the needed info.
Display("Incorporating indexed data for training...");
contexts = di.GetContexts();
values = di.Values;
cutoff = modelCutoff;
predicateCounts = di.GetPredCounts();
numTimesEventsSeen = di.GetNumTimesEventsSeen();
numUniqueEvents = contexts.Length;
prior = modelPrior;
// determine the correction constant and its inverse
double correctionConstant = 0;
for (int ci = 0; ci < contexts.Length; ci++) {
if (values == null || values[ci] == null) {
if (contexts[ci].Length > correctionConstant) {
correctionConstant = contexts[ci].Length;
}
} else {
var cl = values[ci][0];
for (var vi = 1; vi < values[ci].Length; vi++) {
cl += values[ci][vi];
}
if (cl > correctionConstant) {
correctionConstant = cl;
}
}
}
Display("done.");
outcomeLabels = di.GetOutcomeLabels();
outcomeList = di.GetOutcomeList();
numOutcomes = outcomeLabels.Length;
predLabels = di.GetPredLabels();
prior.SetLabels(outcomeLabels, predLabels);
numPreds = predLabels.Length;
info.Append("Number of Event Tokens: {0}\n", numUniqueEvents);
info.Append(" Number of Outcomes: {0}\n", numOutcomes);
info.Append(" Number of Predicates: {0}\n", numPreds);
Display("\tNumber of Event Tokens: " + numUniqueEvents);
Display("\t Number of Outcomes: " + numOutcomes);
Display("\t Number of Predicates: " + numPreds);
// set up feature arrays
//var predCount = new float[numPreds][numOutcomes];
var predCount = new float[numPreds][];
for (int ti = 0; ti < numUniqueEvents; ti++) {
for (int j = 0; j < contexts[ti].Length; j++) {
if (predCount[contexts[ti][j]] == null) {
predCount[contexts[ti][j]] = new float[numOutcomes];
}
if (values != null && values[ti] != null) {
predCount[contexts[ti][j]][outcomeList[ti]] += numTimesEventsSeen[ti]*values[ti][j];
} else {
predCount[contexts[ti][j]][outcomeList[ti]] += numTimesEventsSeen[ti];
}
}
}
// ReSharper disable once RedundantAssignment
di = null;
// Get the observed expectations of the features. Strictly speaking,
// we should divide the counts by the number of Tokens, but because of
// the way the model's expectations are approximated in the
// implementation, this is canceled out when we compute the next
// iteration of a parameter, making the extra divisions wasteful.
param = new MutableContext[numPreds];
for (var i = 0; i < modelExpects.Length; i++)
modelExpects[i] = new MutableContext[numPreds];
observedExpects = new MutableContext[numPreds];
// The model does need the correction constant and the correction feature. The correction constant
// is only needed during training, and the correction feature is not necessary.
// For compatibility reasons the model contains form now on a correction constant of 1,
// and a correction param 0.
// ReSharper disable once CoVariantArrayConversion
evalParams = new EvalParameters(param, 0, 1, numOutcomes);
var activeOutcomes = new int[numOutcomes];
var allOutcomesPattern = new int[numOutcomes];
for (var oi = 0; oi < numOutcomes; oi++) {
allOutcomesPattern[oi] = oi;
}
for (var pi = 0; pi < numPreds; pi++) {
var numActiveOutcomes = 0;
int[] outcomePattern;
if (Smoothing) {
numActiveOutcomes = numOutcomes;
outcomePattern = allOutcomesPattern;
} else {
//determine active outcomes
for (var oi = 0; oi < numOutcomes; oi++) {
if (predCount[pi][oi] > 0 && predicateCounts[pi] >= cutoff) {
activeOutcomes[numActiveOutcomes] = oi;
numActiveOutcomes++;
}
}
if (numActiveOutcomes == numOutcomes) {
outcomePattern = allOutcomesPattern;
} else {
outcomePattern = new int[numActiveOutcomes];
for (var aoi = 0; aoi < numActiveOutcomes; aoi++) {
outcomePattern[aoi] = activeOutcomes[aoi];
}
}
}
param[pi] = new MutableContext(outcomePattern, new double[numActiveOutcomes]);
foreach (MutableContext[] me in modelExpects)
me[pi] = new MutableContext(outcomePattern, new double[numActiveOutcomes]);
observedExpects[pi] = new MutableContext(outcomePattern, new double[numActiveOutcomes]);
for (var aoi = 0; aoi < numActiveOutcomes; aoi++) {
var oi = outcomePattern[aoi];
param[pi].SetParameter(aoi, 0.0);
foreach (var modelExpect in modelExpects) {
modelExpect[pi].SetParameter(aoi, 0.0);
}
if (predCount[pi][oi] > 0) {
observedExpects[pi].SetParameter(aoi, predCount[pi][oi]);
} else if (Smoothing) {
observedExpects[pi].SetParameter(aoi, SmoothingObservation);
}
}
}
Display("...done.");
/***************** Find the parameters ************************/
if (threads == 1)
Display("Computing model parameters ...");
else
Display("Computing model parameters in " + threads + " threads...");
FindParameters(iterations, correctionConstant);
/*************** Create and return the model ******************/
// To be compatible with old models the correction constant is always 1
// ReSharper disable once CoVariantArrayConversion
return new GISModel(param, predLabels, outcomeLabels, 1, evalParams.CorrectionParam) {
info = TrainingInfo
};
}
/// <summary>
/// Train a model using the GIS algorithm.
/// </summary>
/// <param name="iterations">The number of GIS iterations to perform.</param>
/// <param name="di">The data indexer used to compress events in memory.</param>
/// <param name="modelPrior">The prior distribution used to train this model.</param>
/// <param name="modelCutoff">The number of times a feature must occur to be included.</param>
/// <param name="threads">The number of threads used to train this model.</param>
/// <returns>The newly trained model, which can be used immediately or saved to disk using an <see cref="GISModelWriter"/> object.</returns>
public GISModel TrainModel(int iterations, IDataIndexer di, IPrior modelPrior, int modelCutoff, int threads)
{
if (threads <= 0)
{
throw new ArgumentOutOfRangeException("threads", threads, @"Threads must be at least one or greater.");
}
modelExpects = new MutableContext[threads][];
info.Append("Trained using GIS algorithm.\n\n");
// Executes the data indexer
di.Execute();
// Incorporate all of the needed info.
Display("Incorporating indexed data for training...");
contexts = di.GetContexts();
values = di.Values;
cutoff = modelCutoff;
predicateCounts = di.GetPredCounts();
numTimesEventsSeen = di.GetNumTimesEventsSeen();
numUniqueEvents = contexts.Length;
prior = modelPrior;
// determine the correction constant and its inverse
double correctionConstant = 0;
for (int ci = 0; ci < contexts.Length; ci++)
{
if (values == null || values[ci] == null)
{
if (contexts[ci].Length > correctionConstant)
{
correctionConstant = contexts[ci].Length;
}
}
else
{
var cl = values[ci][0];
for (var vi = 1; vi < values[ci].Length; vi++)
{
cl += values[ci][vi];
}
if (cl > correctionConstant)
{
correctionConstant = cl;
}
}
}
Display("done.");
outcomeLabels = di.GetOutcomeLabels();
outcomeList = di.GetOutcomeList();
numOutcomes = outcomeLabels.Length;
predLabels = di.GetPredLabels();
prior.SetLabels(outcomeLabels, predLabels);
numPreds = predLabels.Length;
info.Append("Number of Event Tokens: {0}\n", numUniqueEvents);
info.Append(" Number of Outcomes: {0}\n", numOutcomes);
info.Append(" Number of Predicates: {0}\n", numPreds);
Display("\tNumber of Event Tokens: " + numUniqueEvents);
Display("\t Number of Outcomes: " + numOutcomes);
Display("\t Number of Predicates: " + numPreds);
// set up feature arrays
//var predCount = new float[numPreds][numOutcomes];
var predCount = new float[numPreds][];
for (int ti = 0; ti < numUniqueEvents; ti++)
{
for (int j = 0; j < contexts[ti].Length; j++)
{
if (predCount[contexts[ti][j]] == null)
{
predCount[contexts[ti][j]] = new float[numOutcomes];
}
if (values != null && values[ti] != null)
{
predCount[contexts[ti][j]][outcomeList[ti]] += numTimesEventsSeen[ti] * values[ti][j];
}
else
{
predCount[contexts[ti][j]][outcomeList[ti]] += numTimesEventsSeen[ti];
}
}
}
// ReSharper disable once RedundantAssignment
di = null;
// Get the observed expectations of the features. Strictly speaking,
// we should divide the counts by the number of Tokens, but because of
// the way the model's expectations are approximated in the
// implementation, this is canceled out when we compute the next
// iteration of a parameter, making the extra divisions wasteful.
param = new MutableContext[numPreds];
for (var i = 0; i < modelExpects.Length; i++)
{
modelExpects[i] = new MutableContext[numPreds];
}
observedExpects = new MutableContext[numPreds];
// The model does need the correction constant and the correction feature. The correction constant
// is only needed during training, and the correction feature is not necessary.
// For compatibility reasons the model contains form now on a correction constant of 1,
// and a correction param 0.
// ReSharper disable once CoVariantArrayConversion
evalParams = new EvalParameters(param, 0, 1, numOutcomes);
var activeOutcomes = new int[numOutcomes];
var allOutcomesPattern = new int[numOutcomes];
for (var oi = 0; oi < numOutcomes; oi++)
{
allOutcomesPattern[oi] = oi;
}
for (var pi = 0; pi < numPreds; pi++)
{
var numActiveOutcomes = 0;
int[] outcomePattern;
if (Smoothing)
{
numActiveOutcomes = numOutcomes;
outcomePattern = allOutcomesPattern;
}
else
{
//determine active outcomes
for (var oi = 0; oi < numOutcomes; oi++)
{
if (predCount[pi][oi] > 0 && predicateCounts[pi] >= cutoff)
{
activeOutcomes[numActiveOutcomes] = oi;
numActiveOutcomes++;
}
}
if (numActiveOutcomes == numOutcomes)
{
outcomePattern = allOutcomesPattern;
}
else
{
outcomePattern = new int[numActiveOutcomes];
for (var aoi = 0; aoi < numActiveOutcomes; aoi++)
{
outcomePattern[aoi] = activeOutcomes[aoi];
}
}
}
param[pi] = new MutableContext(outcomePattern, new double[numActiveOutcomes]);
foreach (MutableContext[] me in modelExpects)
{
me[pi] = new MutableContext(outcomePattern, new double[numActiveOutcomes]);
}
observedExpects[pi] = new MutableContext(outcomePattern, new double[numActiveOutcomes]);
for (var aoi = 0; aoi < numActiveOutcomes; aoi++)
{
var oi = outcomePattern[aoi];
param[pi].SetParameter(aoi, 0.0);
foreach (var modelExpect in modelExpects)
{
modelExpect[pi].SetParameter(aoi, 0.0);
}
if (predCount[pi][oi] > 0)
{
observedExpects[pi].SetParameter(aoi, predCount[pi][oi]);
}
else if (Smoothing)
{
observedExpects[pi].SetParameter(aoi, SmoothingObservation);
}
}
}
Display("...done.");
/***************** Find the parameters ************************/
if (threads == 1)
{
Display("Computing model parameters ...");
}
else
{
Display("Computing model parameters in " + threads + " threads...");
}
FindParameters(iterations, correctionConstant);
/*************** Create and return the model ******************/
// To be compatible with old models the correction constant is always 1
// ReSharper disable once CoVariantArrayConversion
return(new GISModel(param, predLabels, outcomeLabels, 1, evalParams.CorrectionParam)
{
info = TrainingInfo
});
}
/// <summary>
/// Execute the training operation.
/// </summary>
/// <param name="iterations">The number of iterations.</param>
/// <param name="indexer">The data indexer.</param>
/// <returns>The trained <see cref="IMaxentModel" /> model.</returns>
/// <exception cref="System.ArgumentOutOfRangeException">iterations</exception>
/// <exception cref="System.ArgumentNullException">indexer</exception>
/// <exception cref="System.InvalidOperationException">The number of threads is invalid.</exception>
public QNModel TrainModel(int iterations, IDataIndexer indexer)
{
if (iterations < 0)
{
throw new ArgumentOutOfRangeException(nameof(iterations));
}
if (indexer == null)
{
throw new ArgumentNullException(nameof(indexer));
}
IFunction function;
if (threads == 1)
{
Display("Computing model parameters ...");
function = new NegLogLikelihood(indexer);
}
else if (threads > 1)
{
Display("Computing model parameters in " + threads + " threads ...");
function = new ParallelNegLogLikelihood(indexer, threads);
}
else
{
throw new InvalidOperationException("The number of threads is invalid.");
}
if (!indexer.Completed)
{
indexer.Execute();
}
var minimizer = new QNMinimizer(l1Cost, l2Cost, iterations, updates, maxFctEval, Monitor)
{
Evaluator = new QNModelEvaluator(indexer)
};
// minimized parameters
var mp = minimizer.Minimize(function);
// construct model with trained parameters
var predLabels = indexer.GetPredLabels();
var nPredLabels = predLabels.Length;
var outcomeNames = indexer.GetOutcomeLabels();
var nOutcomes = outcomeNames.Length;
var parameters = new Context[nPredLabels];
for (var ci = 0; ci < parameters.Length; ci++)
{
var outcomePattern = new List <int>(nOutcomes);
var alpha = new List <double>(nOutcomes);
for (var oi = 0; oi < nOutcomes; oi++)
{
var val = mp[oi * nPredLabels + ci];
outcomePattern.Add(oi);
alpha.Add(val);
}
parameters[ci] = new Context(outcomePattern.ToArray(), alpha.ToArray());
}
return(new QNModel(parameters, predLabels, outcomeNames));
}
/// <summary>
/// Train a model using the Perceptron algorithm.
/// </summary>
/// <param name="iterations">The number of Perceptron iterations to perform.</param>
/// <param name="indexer">The object which will be used for event compilation.</param>
/// <param name="cutoff">The number of times a predicate must occur to be used in a model.</param>
/// <param name="useAverage"></param>
/// <returns>The newly trained model, which can be used immediately or saved to disk using a <see cref="IO.PerceptronModelWriter"/> object.</returns>
public AbstractModel TrainModel(int iterations, IDataIndexer indexer, int cutoff, bool useAverage) {
Display("Incorporating indexed data for training...");
info.Append("Trained using Perceptron algorithm.\n\n");
// Executes the data indexer
indexer.Execute();
contexts = indexer.GetContexts();
values = indexer.Values;
numTimesEventsSeen = indexer.GetNumTimesEventsSeen();
numEvents = indexer.GetNumEvents();
numUniqueEvents = contexts.Length;
outcomeLabels = indexer.GetOutcomeLabels();
outcomeList = indexer.GetOutcomeList();
predLabels = indexer.GetPredLabels();
numPreds = predLabels.Length;
numOutcomes = outcomeLabels.Length;
Display("\ndone.\n");
info.Append("Number of Event Tokens: {0}\n" +
" Number of Outcomes: {1}\n" +
" Number of Predicates: {2}\n", numEvents, numOutcomes, numPreds);
Display("\tNumber of Event Tokens: " + numUniqueEvents);
Display("\t Number of Outcomes: " + numOutcomes);
Display("\t Number of Predicates: " + numPreds);
Display("Computing model parameters.");
var finalParameters = FindParameters(iterations, useAverage);
Display("\ndone.\n");
// ReSharper disable once CoVariantArrayConversion
return new PerceptronModel(finalParameters, predLabels, outcomeLabels) {
info = info
};
}