/// <summary>
/// Sets the traits, biases and thresholds to the product of two parameter distributions.
/// </summary>
/// <param name="firstFactor">The first factor.</param>
/// <param name="secondFactor">The second factor.</param>
/// <remarks>This method does not affect the user and item features.</remarks>
public void SetEntityParametersToProduct(ParameterDistributions firstFactor, ParameterDistributions secondFactor)
{
this.UserTraitDistribution.SetToProduct(firstFactor.UserTraitDistribution, secondFactor.UserTraitDistribution);
this.UserBiasDistribution.SetToProduct(firstFactor.UserBiasDistribution, secondFactor.UserBiasDistribution);
this.UserThresholdDistribution.SetToProduct(firstFactor.UserThresholdDistribution, secondFactor.UserThresholdDistribution);
this.ItemTraitDistribution.SetToProduct(firstFactor.ItemTraitDistribution, secondFactor.ItemTraitDistribution);
this.ItemBiasDistribution.SetToProduct(firstFactor.ItemBiasDistribution, secondFactor.ItemBiasDistribution);
}
/// <summary>
/// Sets the traits, biases and thresholds to the power of a parameter distributions.
/// </summary>
/// <param name="value">The base.</param>
/// <param name="exponent">The exponent.</param>
/// <remarks>This method does not affect the user and item features.</remarks>
public void SetEntityParametersToPower(ParameterDistributions value, double exponent)
{
this.UserTraitDistribution.SetToPower(value.UserTraitDistribution, exponent);
this.UserBiasDistribution.SetToPower(value.UserBiasDistribution, exponent);
this.UserThresholdDistribution.SetToPower(value.UserThresholdDistribution, exponent);
this.ItemTraitDistribution.SetToPower(value.ItemTraitDistribution, exponent);
this.ItemBiasDistribution.SetToPower(value.ItemBiasDistribution, exponent);
}
/// <summary>
/// Sets the traits, biases and thresholds to represent the ratio of two parameter distributions.
/// </summary>
/// <param name="numerator">The numerator parameter distribution.</param>
/// <param name="denominator">The denominator parameter distribution.</param>
/// <param name="forceProper">A flag indicating whether or not to force proper messages.</param>
/// <remarks>This method does not affect the user and item features.</remarks>
public void SetEntityParametersToRatio(ParameterDistributions numerator, ParameterDistributions denominator, bool forceProper = true)
{
this.UserTraitDistribution.SetToRatio(numerator.UserTraitDistribution, denominator.UserTraitDistribution, forceProper);
this.UserBiasDistribution.SetToRatio(numerator.UserBiasDistribution, denominator.UserBiasDistribution, forceProper);
this.UserThresholdDistribution.SetToRatio(numerator.UserThresholdDistribution, denominator.UserThresholdDistribution, forceProper);
this.ItemTraitDistribution.SetToRatio(numerator.ItemTraitDistribution, denominator.ItemTraitDistribution, forceProper);
this.ItemBiasDistribution.SetToRatio(numerator.ItemBiasDistribution, denominator.ItemBiasDistribution, forceProper);
}
/// <summary>
/// Initializes the per-user and per-item model parameters.
/// </summary>
/// <param name="initializers">The parameters to initialize to.</param>
/// <remarks>
/// The per-entity parameter initialization is required for correct reconstruction of the
/// intermediate messages by the special first iteration. The initializers have to contain
/// the corresponding posteriors from the last iteration. Note that feature-related parameters
/// do not need to be initialized, because messages to the features are not reconstructed by
/// the special first iteration.
/// </remarks>
public void InitializeEntityParameters(ParameterDistributions initializers)
{
this.inferenceAlgorithm.UserTraitsInitializer = initializers.UserTraitDistribution;
this.inferenceAlgorithm.UserBiasInitializer = initializers.UserBiasDistribution;
this.inferenceAlgorithm.UserThresholdsInitializer = initializers.UserThresholdDistribution;
this.inferenceAlgorithm.ItemTraitsInitializer = initializers.ItemTraitDistribution;
this.inferenceAlgorithm.ItemBiasInitializer = initializers.ItemBiasDistribution;
}
/// <summary>
/// Constrains the per-user and per-item model parameters.
/// </summary>
/// <param name="constraints">The parameters to constrain to.</param>
/// <remarks>
/// The per-entity parameters are constrained to their corresponding output messages
/// (or more concretely, to the marginal divided by the prior). This is required in
/// order to carry message information over several batches of the data.
/// </remarks>
public void ConstrainEntityParameters(ParameterDistributions constraints)
{
this.inferenceAlgorithm.UserTraitsMessage = constraints.UserTraitDistribution;
this.inferenceAlgorithm.UserBiasMessage = constraints.UserBiasDistribution;
this.inferenceAlgorithm.UserThresholdsMessage = constraints.UserThresholdDistribution;
this.inferenceAlgorithm.ItemTraitsMessage = constraints.ItemTraitDistribution;
this.inferenceAlgorithm.ItemBiasMessage = constraints.ItemBiasDistribution;
}
/// <summary>
/// Gets the parameter distributions of the average item.
/// </summary>
/// <param name="parameterDistributions">The learned posterior parameter distributions.</param>
/// <param name="itemFeatures">The item features.</param>
/// <returns>The average item parameter distributions.</returns>
public static ItemParameterDistribution GetAverageItemParameters(
ParameterDistributions parameterDistributions, SparseFeatureMatrix itemFeatures)
{
GaussianArray traitDistribution;
Gaussian biasDistribution;
GetAverageTraitsAndBiases(
parameterDistributions.ItemTraitDistribution,
parameterDistributions.ItemBiasDistribution,
parameterDistributions.ItemFeature.TraitWeights,
parameterDistributions.ItemFeature.BiasWeights,
itemFeatures,
out traitDistribution,
out biasDistribution);
return(new ItemParameterDistribution(traitDistribution, biasDistribution));
}
/// <summary>
/// Gets the parameter distributions of the average user.
/// </summary>
/// <param name="parameterDistributions">The learned posterior parameter distributions.</param>
/// <param name="userFeatures">The user features.</param>
/// <returns>The average user parameter distributions.</returns>
public static UserParameterDistribution GetAverageUserParameters(
ParameterDistributions parameterDistributions, SparseFeatureMatrix userFeatures)
{
GaussianArray traitDistribution;
Gaussian biasDistribution;
GaussianArray thresholdDistribution = GetAverageParameterVector(parameterDistributions.UserThresholdDistribution);
GetAverageTraitsAndBiases(
parameterDistributions.UserTraitDistribution,
parameterDistributions.UserBiasDistribution,
parameterDistributions.UserFeature.TraitWeights,
parameterDistributions.UserFeature.BiasWeights,
userFeatures,
out traitDistribution,
out biasDistribution);
return(new UserParameterDistribution(traitDistribution, biasDistribution, thresholdDistribution));
}