File prediction_complete.hpp

template<typename ScoreType>
class CompletePrediction : public VectorDecorator<AllocatedVector<ScoreType>>, public IEvaluatedPrediction

#include <prediction_complete.hpp>

Stores the scores that are predicted by a rule that predicts for all available outputs.

Template Parameters:

ScoreType – The type of the predicted scores

Public Types

using value_iterator = View<ScoreType>::iterator

An iterator that provides access to the predicted scores and allows to modify them.

using value_const_iterator = View<ScoreType>::const_iterator

An iterator that provides read-only access to the predicted scores.

using index_const_iterator = CompleteIndexVector::const_iterator

An iterator that provides read-only access to the indices of the outputs for which the rule predicts.

Public Functions

CompletePrediction(uint32 numElements, IStatisticsUpdateFactory<ScoreType> &statisticsUpdateFactory)

Parameters:

• numElements – The number of outputs for which the rule predicts

• statisticsUpdateFactory – A reference to an object of type IStatisticsUpdateFactory

value_iterator values_begin()

Returns a value_iterator to the beginning of the predicted scores.

Returns:

A value_iterator to the beginning

value_iterator values_end()

Returns a value_iterator to the end of the predicted scores.

Returns:

A value_iterator to the end

value_const_iterator values_cbegin() const

Returns a value_const_iterator to the beginning of the predicted scores.

Returns:

A value_const_iterator to the beginning

value_const_iterator values_cend() const

Returns a const_iterator to the end of the predicted scores.

Returns:

A const_iterator to the end

index_const_iterator indices_cbegin() const

Returns an index_const_iterator to the beginning of the indices of the outputs for which the rule predicts.

Returns:

An index_const_iterator to the beginning

index_const_iterator indices_cend() const

Returns an index_const_iterator to the end of the indices of the outputs for which the rule predicts.

Returns:

An index_const_iterator to the end

virtual uint32 getNumElements() const override

Returns the number of indices.

Returns:

The number of indices

virtual void sort() override

Sorts the scores that are stored by this prediction in increasing order by the indices of the outputs they correspond to.

virtual void postProcess(const IPostProcessor &postProcessor) override

Post-processes the scores that are stored by this prediction.

Parameters:

postProcessor – A reference to an object of type IPostProcessor that should be used for post-processing

virtual bool isPartial() const override

Returns whether the indices are partial, i.e., some indices in the range [0, getNumElements()) are missing, or not.

Returns:

True, if the indices are partial, false otherwise

virtual uint32 getIndex(uint32 pos) const override

Returns the index at a specific position.

Parameters:

pos – The position of the index. Must be in [0, getNumElements())

Returns:

The index at the given position

virtual void visit(PartialIndexVectorVisitor partialIndexVectorVisitor, CompleteIndexVectorVisitor completeIndexVectorVisitor) const override

Invokes one of the given visitor functions, depending on which one is able to handle this particular type of vector.

Parameters:

• partialIndexVectorVisitor – The visitor function for handling objects of the type PartialIndexVector

• completeIndexVectorVisitor – The visitor function for handling objects of the type CompleteIndexVector

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const EqualWeightVector &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type EqualWeightVector that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const BitWeightVector &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type BitWeightVector that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const DenseWeightVector<uint16> &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type DenseWeightVector<uint16> that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const DenseWeightVector<float32> &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type DenseWeightVector<float32> that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const OutOfSampleWeightVector<EqualWeightVector> &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type OutOfSampleWeightVector<EqualWeightVector> that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const OutOfSampleWeightVector<BitWeightVector> &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type OutOfSampleWeightVector<BitWeightVector> that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const OutOfSampleWeightVector<DenseWeightVector<uint16>> &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type OutOfSampleWeightVector<DenseWeightVector<uint16>> that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual std::unique_ptr<IStatisticsSubset> createStatisticsSubset(const IStatistics &statistics, const OutOfSampleWeightVector<DenseWeightVector<float32>> &weights) const override

Creates and returns a new subset of the given statistics that only contains the outputs whose indices are stored in this vector.

Parameters:

• statistics – A reference to an object of type IStatistics that should be used to create the subset

• weights – A reference to an object of type OutOfSampleWeightVector<DenseWeightVector<float32>> that provides access to the weights of individual training examples

Returns:

An unique pointer to an object of type IStatisticsSubset that has been created

virtual void applyPrediction(uint32 statisticIndex) override

Updates a specific statistic.

This function must be called for each statistic that is covered by a new rule before learning the next rule.

Parameters:

statisticIndex – The index of the statistic that should be updated

virtual void revertPrediction(uint32 statisticIndex) override

Reverts a specific statistic that has previously been updated via the function applyPrediction.

Parameters:

statisticIndex – The index of the statistic that should be updated

virtual std::unique_ptr<IHead> createHead() const override

Creates and returns a head that contains the scores that are stored by this prediction.

Returns:

An unique pointer to an object of type IHead that has been created

Private Members

const CompleteIndexVector indexVector_

const std::unique_ptr<IStatisticsUpdate> statisticsUpdatePtr_