org.apache.spark.mllib.linalg

Class EigenValueDecomposition

Object

org.apache.spark.mllib.linalg.EigenValueDecomposition

public class EigenValueDecomposition
extends Object

Compute eigen-decomposition.

Constructor Summary

Constructors

Constructor and Description
EigenValueDecomposition()

Method Summary

Modifier and Type	Method and Description
static scala.Tuple2<breeze.linalg.DenseVector<Object>,breeze.linalg.DenseMatrix<Object>>	symmetricEigs(scala.Function1<breeze.linalg.DenseVector<Object>,breeze.linalg.DenseVector<Object>> mul, int n, int k, double tol, int maxIterations) Compute the leading k eigenvalues and eigenvectors on a symmetric square matrix using ARPACK.

Methods inherited from class Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

EigenValueDecomposition

public EigenValueDecomposition()

Method Detail

symmetricEigs

public static scala.Tuple2<breeze.linalg.DenseVector<Object>,breeze.linalg.DenseMatrix<Object>> symmetricEigs(scala.Function1<breeze.linalg.DenseVector<Object>,breeze.linalg.DenseVector<Object>> mul,
                                                                                                              int n,
                                                                                                              int k,
                                                                                                              double tol,
                                                                                                              int maxIterations)

Compute the leading k eigenvalues and eigenvectors on a symmetric square matrix using ARPACK. The caller needs to ensure that the input matrix is real symmetric. This function requires memory for n*(4*k+4) doubles.

Parameters:

mul - a function that multiplies the symmetric matrix with a DenseVector.

n - dimension of the square matrix (maximum Int.MaxValue).

k - number of leading eigenvalues required, where k must be positive and less than n.

tol - tolerance of the eigs computation.

maxIterations - the maximum number of Arnoldi update iterations.

Returns:

a dense vector of eigenvalues in descending order and a dense matrix of eigenvectors (columns of the matrix).

Note:

The number of computed eigenvalues might be smaller than k when some Ritz values do not satisfy the convergence criterion specified by tol (see ARPACK Users Guide, Chapter 4.6 for more details). The maximum number of Arnoldi update iterations is set to 300 in this function.