Java Code Examples for org.apache.commons.math3.linear.CholeskyDecomposition#getL()

/**
 * Calculates the Cholesky decomposition of a matrix. The Cholesky
 * decomposition of a real symmetric positive-definite matrix A consists of
 * a lower triangular matrix L with same size such that: A = LLT. In a
 * sense, this is the square root of A.
 *
 * @param a The given matrix.
 * @return Result array.
 */
public static Array cholesky(Array a) {
    Array r = Array.factory(DataType.DOUBLE, a.getShape());
    double[][] aa = (double[][]) ArrayUtil.copyToNDJavaArray_Double(a);
    RealMatrix matrix = new Array2DRowRealMatrix(aa, false);
    CholeskyDecomposition decomposition = new CholeskyDecomposition(matrix);
    RealMatrix L = decomposition.getL();
    int n = L.getColumnDimension();
    int m = L.getRowDimension();
    for (int i = 0; i < m; i++) {
        for (int j = 0; j < n; j++) {
            r.setDouble(i * n + j, L.getEntry(i, j));
        }
    }

    return r;
}
 

/**
 * Function to compute Cholesky decomposition of the given input matrix. 
 * The input must be a real symmetric positive-definite matrix.
 * 
 * @param in commons-math3 Array2DRowRealMatrix
 * @return matrix block
 */
private static MatrixBlock computeCholesky(Array2DRowRealMatrix in) {
	if ( !in.isSquare() )
		throw new DMLRuntimeException("Input to cholesky() must be square matrix -- given: a " + in.getRowDimension() + "x" + in.getColumnDimension() + " matrix.");
	CholeskyDecomposition cholesky = new CholeskyDecomposition(in, 1e-14,
		CholeskyDecomposition.DEFAULT_ABSOLUTE_POSITIVITY_THRESHOLD);
	RealMatrix rmL = cholesky.getL();
	return DataConverter.convertToMatrixBlock(rmL.getData());
}
 

public void fit(List<Location> locations){
	setLocations(locations);
	int n = locations.size();
	double[][] K = new double[n][n];

	for(int i=0; i<n; i++){
		Location loc1 = locations.get(i);
		double[] x1 = ModelAdaptUtils.locationToVec(loc1);
		for(int j=i; j<n; j++){
			Location loc2 = locations.get(j);
			double[] x2 = ModelAdaptUtils.locationToVec(loc2);
			double k =kernel.computeKernel(x1, x2);
			K[i][j] = k;
			K[j][i] = k;
		}
	}
	RealMatrix Kmat = MatrixUtils.createRealMatrix(K);
	RealMatrix lambdamat = MatrixUtils.createRealIdentityMatrix(n).scalarMultiply(sigma_n*sigma_n); //Tentative treatment

	RealMatrix Kymat = Kmat.add(lambdamat);

	CholeskyDecomposition chol = new CholeskyDecomposition(Kymat);
	RealMatrix Lmat = chol.getL();

	double[] normalRands = new double[n];
	for(int i=0; i<n; i++){
		normalRands[i] = rand.nextGaussian();
	}
	this.y = Lmat.operate(normalRands);

	RealMatrix invKymat = (new LUDecomposition(Kymat)).getSolver().getInverse();
	this.alphas = invKymat.operate(y);
}
 

/**
 * Function to compute Cholesky decomposition of the given input matrix. 
 * The input must be a real symmetric positive-definite matrix.
 * 
 * @param in commons-math3 Array2DRowRealMatrix
 * @return matrix block
 */
private static MatrixBlock computeCholesky(Array2DRowRealMatrix in) {
	if ( !in.isSquare() )
		throw new DMLRuntimeException("Input to cholesky() must be square matrix -- given: a " + in.getRowDimension() + "x" + in.getColumnDimension() + " matrix.");
	CholeskyDecomposition cholesky = new CholeskyDecomposition(in, 1e-14,
		CholeskyDecomposition.DEFAULT_ABSOLUTE_POSITIVITY_THRESHOLD);
	RealMatrix rmL = cholesky.getL();
	return DataConverter.convertToMatrixBlock(rmL.getData());
}
 

void setCovariance(RealMatrix covariance) {
    this.covariance = covariance;
    CholeskyDecomposition decomposition = new CholeskyDecomposition(covariance);
    this.inverseCovariance = decomposition.getSolver().getInverse();
    RealMatrix lMatrix = decomposition.getL();
    double sum = 0;
    for (int i=0;i<lMatrix.getRowDimension();i++){
        sum += Math.log(lMatrix.getEntry(i,i));
    }
    this.logDeterminant = 2*sum;
}