/// <summary>
/// Use this model to evaluate a context and return an array of the likelihood
/// of each outcome given that context.
/// </summary>
/// <param name="context">
/// The names of the predicates which have been observed at the
/// present decision point. </param>
/// <param name="outsums">
/// This is where the distribution is stored. </param>
/// <returns> The normalized probabilities for the outcomes given the context.
/// The indexes of the double[] are the outcome ids, and the actual
/// string representation of the outcomes can be obtained from the
/// method getOutcome(int i). </returns>
public double[] eval(string[] context, float[] values, double[] outsums)
{
int[] scontexts = new int[context.Length];
for (int i = 0; i < context.Length; i++)
{
int?ci = pmap.get(context[i]);
scontexts[i] = !ci.HasValue ? -1 : ci.GetValueOrDefault();
}
prior.logPrior(outsums, scontexts, values);
return(GISModel.eval(scontexts, values, outsums, evalParams));
}
public virtual ModelExpactationComputeTask call()
{
double[] modelDistribution = new double[outerInstance.numOutcomes];
for (int ei = startIndex; ei < startIndex + length; ei++)
{
// TODO: check interruption status here, if interrupted set a poisoned flag and return
if (outerInstance.values != null)
{
outerInstance.prior.logPrior(modelDistribution, outerInstance.contexts[ei],
outerInstance.values[ei]);
GISModel.eval(outerInstance.contexts[ei], outerInstance.values[ei], modelDistribution,
outerInstance.evalParams);
}
else
{
outerInstance.prior.logPrior(modelDistribution, outerInstance.contexts[ei]);
GISModel.eval(outerInstance.contexts[ei], modelDistribution, outerInstance.evalParams);
}
for (int j = 0; j < outerInstance.contexts[ei].Length; j++)
{
int pi = outerInstance.contexts[ei][j];
if (outerInstance.predicateCounts[pi] >= outerInstance.cutoff)
{
int[] activeOutcomes = outerInstance.modelExpects[threadIndex][pi].Outcomes;
for (int aoi = 0; aoi < activeOutcomes.Length; aoi++)
{
int oi = activeOutcomes[aoi];
// numTimesEventsSeen must also be thread safe
if (outerInstance.values != null && outerInstance.values[ei] != null)
{
outerInstance.modelExpects[threadIndex][pi].updateParameter(aoi,
modelDistribution[oi] * outerInstance.values[ei][j] *
outerInstance.numTimesEventsSeen[ei]);
}
else
{
outerInstance.modelExpects[threadIndex][pi].updateParameter(aoi,
modelDistribution[oi] * outerInstance.numTimesEventsSeen[ei]);
}
}
}
}
loglikelihood += Math.Log(modelDistribution[outerInstance.outcomeList[ei]]) *
outerInstance.numTimesEventsSeen[ei];
numEvents += outerInstance.numTimesEventsSeen[ei];
if (outerInstance.printMessages)
{
int max = 0;
for (int oi = 1; oi < outerInstance.numOutcomes; oi++)
{
if (modelDistribution[oi] > modelDistribution[max])
{
max = oi;
}
}
if (max == outerInstance.outcomeList[ei])
{
numCorrect += outerInstance.numTimesEventsSeen[ei];
}
}
}
return(this);
}
