/// <summary>
/// Priors from posteriors.
/// </summary>
/// <param name="posteriors">The posteriors.</param>
/// <param name="noiseVariance">The noise variance.</param>
/// <returns>
/// The <see cref="Priors" />
/// </returns>
internal static Priors FromPosteriors(Posteriors posteriors, double noiseVariance)
{
return(new Priors
{
Weights = posteriors.Weights,
Threshold = posteriors.Threshold,
NoiseVariance = noiseVariance
});
}
/// <summary>
/// Priors from posteriors with possible missing features.
/// </summary>
/// <param name="posteriors">The posteriors.</param>
/// <param name="featureSet">The feature set.</param>
/// <param name="noiseVariance">The noise variance.</param>
/// <returns>The <see cref="Priors" /></returns>
internal static Priors FromPosteriors(Posteriors posteriors, FeatureSet featureSet, double noiseVariance)
{
return(new Priors
{
Weights =
featureSet.FeatureBuckets.ToDictionary(
ia => ia,
ia =>
posteriors.Weights.ContainsKey(ia)
? posteriors.Weights[ia]
: Gaussian.FromMeanAndVariance(0.0, noiseVariance)),
Threshold = posteriors.Threshold,
NoiseVariance = noiseVariance,
});
}
/// <summary>
/// Runs the experiment.
/// </summary>
/// <param name="trainModel">The train model.</param>
/// <param name="testModel">The test model.</param>
/// <param name="inputs">The inputs.</param>
/// <param name="priors">The priors.</param>
/// <param name="thresholdAndNoiseVariance">The threshold and noise variance.</param>
/// <param name="limit">The limit.</param>
/// <param name="trainMode">The train mode.</param>
/// <param name="testMode">The test mode.</param>
public void Run(
ReplyToModelBase trainModel,
ReplyToModelBase testModel,
Inputs inputs,
Priors priors,
double thresholdAndNoiseVariance,
int limit = -1,
InputMode trainMode = InputMode.Training,
InputMode testMode = InputMode.Validation | InputMode.Testing)
{
this.ModelName = trainModel.Name;
this.FeatureSetName = inputs.FeatureSet.ToString();
trainModel.ConstructModel();
testModel.ConstructModel();
int numTrain = trainMode == InputMode.Training ? inputs.Train.Count : inputs.TrainAndValidation.Count;
if (limit > 0)
{
numTrain = Math.Min(numTrain, limit);
}
if (this.BatchSize == -1)
{
this.BatchSize = numTrain;
}
int numBatches = 1;
if (this.Mode == ExperimentMode.Incremental || this.Mode == ExperimentMode.Online)
{
numBatches = numTrain / this.BatchSize;
}
for (int index = 0; index < numBatches; index++)
{
var batchResults = new Results();
var batch = this.GetBatch(inputs, index, trainMode);
if (numBatches > 1)
{
Console.WriteLine(@"{0} batch {1}/{2}, batch size {3}", this.UserName, index + 1, numBatches, batch.Count);
}
var stopwatch = new Stopwatch();
stopwatch.Start();
trainModel.SetObservedValues(batch, inputs.FeatureSet, InputMode.Training, priors);
try
{
trainModel.DoInference(inputs.FeatureSet, ref batchResults);
}
catch (Exception e)
{
Console.WriteLine(e.Message);
return;
}
stopwatch.Stop();
this.TrainTimings.Add(stopwatch.Elapsed);
// Save the last set of posteriors
this.Posteriors = batchResults.Posteriors;
// Set the priors to be the posteriors from the previous run
var newPriors = Priors.FromPosteriors(this.Posteriors, inputs.FeatureSet, thresholdAndNoiseVariance);
if (this.Mode == ExperimentMode.Online || this.Mode == ExperimentMode.Incremental)
{
if (this.Mode == ExperimentMode.Online)
{
priors = newPriors;
}
// clear posteriors to save on memory
batchResults.Posteriors = null;
}
stopwatch.Restart();
// Validation and Test only
testModel.Apply(inputs, inputs.FeatureSet, newPriors, ref batchResults, testMode);
stopwatch.Stop();
this.TestTimings.Add(stopwatch.Elapsed);
this.Results.Add(batchResults);
this.Metrics.Add(new Metrics(inputs, batchResults, this.Mode));
this.Count++;
}
}
