Exemplos de rank_update_vector_i em C++ (Cpp)

Exemplo n.º 1

Arquivo: locally_linear.hpp Projeto: serialhex/tapkee

DenseSymmetricMatrixPair construct_neighborhood_preserving_eigenproblem(SparseWeightMatrix W,
		RandomAccessIterator begin, RandomAccessIterator end, FeatureVectorCallback feature_vector_callback,
		unsigned int dimension)
{
	timed_context context("NPE eigenproblem construction");
	
	DenseSymmetricMatrix lhs = DenseSymmetricMatrix::Zero(dimension,dimension);
	DenseSymmetricMatrix rhs = DenseSymmetricMatrix::Zero(dimension,dimension);

	DenseVector rank_update_vector_i(dimension);
	DenseVector rank_update_vector_j(dimension);
	for (RandomAccessIterator iter=begin; iter!=end; ++iter)
	{
		feature_vector_callback(*iter,rank_update_vector_i);
		rhs.selfadjointView<Eigen::Upper>().rankUpdate(rank_update_vector_i);
	}

	for (int i=0; i<W.outerSize(); ++i)
	{
		for (SparseWeightMatrix::InnerIterator it(W,i); it; ++it)
		{
			feature_vector_callback(begin[it.row()],rank_update_vector_i);
			feature_vector_callback(begin[it.col()],rank_update_vector_j);
			lhs.selfadjointView<Eigen::Upper>().rankUpdate(rank_update_vector_i, rank_update_vector_j, it.value());
		}
	}
	
	rhs += rhs.transpose();
	rhs /= 2;

	return DenseSymmetricMatrixPair(lhs,rhs);
}

Exemplo n.º 2

Arquivo: laplacian_eigenmaps.hpp Projeto: cwidmer/shogun

DenseSymmetricMatrixPair construct_locality_preserving_eigenproblem(SparseWeightMatrix L,
		DenseDiagonalMatrix D, RandomAccessIterator begin, RandomAccessIterator end, FeatureVectorCallback feature_vector_callback,
		unsigned int dimension)
{
	timed_context context("Constructing LPP eigenproblem");

	DenseSymmetricMatrix lhs = DenseSymmetricMatrix::Zero(dimension,dimension);
	DenseSymmetricMatrix rhs = DenseSymmetricMatrix::Zero(dimension,dimension);

	DenseVector rank_update_vector_i(dimension);
	DenseVector rank_update_vector_j(dimension);
	for (RandomAccessIterator iter=begin; iter!=end; ++iter)
	{
		feature_vector_callback(*iter,rank_update_vector_i);
		rhs.selfadjointView<Eigen::Upper>().rankUpdate(rank_update_vector_i,D.diagonal()(iter-begin));
	}

	for (int i=0; i<L.outerSize(); ++i)
	{
		for (SparseWeightMatrix::InnerIterator it(L,i); it; ++it)
		{
			feature_vector_callback(begin[it.row()],rank_update_vector_i);
			feature_vector_callback(begin[it.col()],rank_update_vector_j);
			lhs.selfadjointView<Eigen::Upper>().rankUpdate(rank_update_vector_i, rank_update_vector_j, it.value());
		}
	}

	return DenseSymmetricMatrixPair(lhs,rhs);
}

Exemplo n.º 3

Arquivo: locally_linear.hpp Projeto: 42MachineLearning/shogun

DenseSymmetricMatrixPair construct_lltsa_eigenproblem(SparseWeightMatrix W,
		RandomAccessIterator begin, RandomAccessIterator end, FeatureVectorCallback feature_vector_callback,
		IndexType dimension)
{
	timed_context context("LLTSA eigenproblem construction");

	DenseSymmetricMatrix lhs = DenseSymmetricMatrix::Zero(dimension,dimension);
	DenseSymmetricMatrix rhs = DenseSymmetricMatrix::Zero(dimension,dimension);

	DenseVector rank_update_vector_i(dimension);
	DenseVector rank_update_vector_j(dimension);
	DenseVector sum = DenseVector::Zero(dimension);

	//RESTRICT_ALLOC;
	for (RandomAccessIterator iter=begin; iter!=end; ++iter)
	{
		feature_vector_callback.vector(*iter,rank_update_vector_i);
		sum += rank_update_vector_i;
		rhs.selfadjointView<Eigen::Upper>().rankUpdate(rank_update_vector_i);
	}
	rhs.selfadjointView<Eigen::Upper>().rankUpdate(sum,-1./(end-begin));

	for (int i=0; i<W.outerSize(); ++i)
	{
		for (SparseWeightMatrix::InnerIterator it(W,i); it; ++it)
		{
			feature_vector_callback.vector(begin[it.row()],rank_update_vector_i);
			feature_vector_callback.vector(begin[it.col()],rank_update_vector_j);
			lhs.selfadjointView<Eigen::Upper>().rankUpdate(rank_update_vector_i, rank_update_vector_j, it.value());
		}
	}
	lhs.selfadjointView<Eigen::Upper>().rankUpdate(sum,-1./(end-begin));

	rhs += rhs.transpose().eval();
	rhs /= 2;

	//UNRESTRICT_ALLOC;

	return DenseSymmetricMatrixPair(lhs,rhs);
}