Examples with DMatrixRMaj org.ejml.data.DMatrixRMaj used on opensource projects

Example 31 with DMatrixRMaj

use of org.ejml.data.DMatrixRMaj in project BoofCV by lessthanoptimal.

the class PoseFromPairLinear6 method massageResults.

/**
 * <p>
 * Since the linear solution is probably not an exact rotation matrix, this code finds the best
 * approximation.
 * </p>
 *
 * See page 280 of [1]
 */
private void massageResults() {
    DMatrixRMaj R = motion.getR();
    Vector3D_F64 T = motion.getT();
    if (!svd.decompose(R))
        throw new RuntimeException("SVD Failed");
    CommonOps_DDRM.multTransB(svd.getU(null, false), svd.getV(null, false), R);
    // determinant should be +1
    double det = CommonOps_DDRM.det(R);
    if (det < 0)
        CommonOps_DDRM.scale(-1, R);
    // compute the determinant of the singular matrix
    double b = 1.0;
    double[] s = svd.getSingularValues();
    for (int i = 0; i < svd.numberOfSingularValues(); i++) {
        b *= s[i];
    }
    b = Math.signum(det) / Math.pow(b, 1.0 / 3.0);
    GeometryMath_F64.scale(T, b);
}

Example 32 with DMatrixRMaj

use of org.ejml.data.DMatrixRMaj in project clusterMaker2 by RBVI.

the class EjmlOps method colMean.

public static DMatrixRMaj colMean(DMatrixRMaj y, int i) {
    DMatrixRMaj colmean = new DMatrixRMaj(1, y.numCols);
    sumCols(y, colmean);
    divide(colmean, y.numRows);
    return colmean;
}

Example 33 with DMatrixRMaj

use of org.ejml.data.DMatrixRMaj in project clusterMaker2 by RBVI.

the class EjmlOps method fillWithRow.

public static DMatrixRMaj fillWithRow(DMatrixRMaj matrix, int setrow) {
    int rows = matrix.numRows;
    int cols = matrix.numCols;
    DMatrixRMaj result = new DMatrixRMaj(rows, cols);
    for (int row = 0; row < rows; row++) {
        for (int col = 0; col < cols; col++) {
            result.set(row, col, matrix.get(setrow, col));
        }
    }
    return result;
}

Example 34 with DMatrixRMaj

use of org.ejml.data.DMatrixRMaj in project clusterMaker2 by RBVI.

the class FastTSne method tsne.

// 
@Override
public double[][] tsne(TSneConfiguration config, TaskMonitor monitor) {
    double[][] X = config.getXin();
    int no_dims = config.getOutputDims();
    int initial_dims = config.getInitialDims();
    double perplexity = config.getPerplexity();
    int max_iter = config.getMaxIter();
    boolean use_pca = config.usePca();
    this.monitor = monitor;
    this.config = config;
    String IMPLEMENTATION_NAME = this.getClass().getSimpleName();
    // System.out.println("X:Shape is = " + X.length + " x " + X[0].length);
    monitor.showMessage(TaskMonitor.Level.INFO, "Running " + IMPLEMENTATION_NAME + ".");
    long end = System.currentTimeMillis();
    long start = System.currentTimeMillis();
    long total = System.currentTimeMillis();
    // Initialize variables
    if (use_pca && X[0].length > initial_dims && initial_dims > 0) {
        monitor.showMessage(TaskMonitor.Level.INFO, "Using PCA to reduce dimensions");
        PrincipalComponentAnalysis pca = new PrincipalComponentAnalysis();
        X = pca.pca(X, initial_dims);
        monitor.showMessage(TaskMonitor.Level.INFO, "X:Shape after PCA is = " + X.length + " x " + X[0].length);
    }
    int n = X.length;
    double momentum = .5;
    double initial_momentum = 0.5;
    double final_momentum = 0.8;
    int eta = 500;
    double min_gain = 0.01;
    DMatrixRMaj Y = new DMatrixRMaj(rnorm(n, no_dims));
    // DMatrixRMaj Y        = new DMatrixRMaj(fillMatrix(n,no_dims,.5));
    DMatrixRMaj Ysqlmul = new DMatrixRMaj(Y.numRows, Y.numRows);
    DMatrixRMaj dY = new DMatrixRMaj(fillMatrix(n, no_dims, 0.0));
    DMatrixRMaj iY = new DMatrixRMaj(fillMatrix(n, no_dims, 0.0));
    DMatrixRMaj gains = new DMatrixRMaj(fillMatrix(n, no_dims, 1.0));
    DMatrixRMaj btNeg = new DMatrixRMaj(n, no_dims);
    DMatrixRMaj bt = new DMatrixRMaj(n, no_dims);
    if (config.cancelled())
        return null;
    // Compute P-values
    // P = n x n
    DMatrixRMaj P = new DMatrixRMaj(x2p(X, 1e-5, perplexity).P);
    if (config.cancelled())
        return null;
    DMatrixRMaj Ptr = new DMatrixRMaj(P.numRows, P.numCols);
    // L = n x n
    DMatrixRMaj L = new DMatrixRMaj(P);
    DMatrixRMaj logdivide = new DMatrixRMaj(P.numRows, P.numCols);
    DMatrixRMaj diag = new DMatrixRMaj(fillMatrix(L.numRows, L.numCols, 0.0));
    transpose(P, Ptr);
    addEquals(P, Ptr);
    divide(P, elementSum(P));
    replaceNaN(P, Double.MIN_VALUE);
    // early exaggeration
    scale(4.0, P);
    maximize(P, 1e-12);
    if (config.cancelled())
        return null;
    monitor.showMessage(TaskMonitor.Level.INFO, "Y:Shape is = " + Y.getNumRows() + " x " + Y.getNumCols());
    DMatrixRMaj sqed = new DMatrixRMaj(Y.numRows, Y.numCols);
    DMatrixRMaj sum_Y = new DMatrixRMaj(1, Y.numRows);
    DMatrixRMaj num = new DMatrixRMaj(Y.numRows, Y.numRows);
    DMatrixRMaj Q = new DMatrixRMaj(P.numRows, P.numCols);
    System.out.println("Created sum_Y: (" + "" + sum_Y.numRows + "," + sum_Y.numCols + ")");
    double progress = 0.0;
    for (int iter = 0; iter < max_iter && !abort; iter++) {
        if (config.cancelled())
            return null;
        progress = (double) iter / (double) max_iter;
        monitor.setProgress(progress);
        // Compute pairwise affinities
        elementPower(Y, 2, sqed);
        sumRows(sqed, sum_Y);
        // Transpose the sum vector.  The sumRows now
        // gratuitously tansposes sum_Y
        transpose(sum_Y);
        multAddTransB(-2.0, Y, Y, Ysqlmul);
        addRowVector(Ysqlmul, sum_Y);
        transpose(Ysqlmul);
        addRowVector(Ysqlmul, sum_Y);
        add(Ysqlmul, 1.0);
        divide(1.0, Ysqlmul);
        num.set(Ysqlmul);
        assignAtIndex(num, range(n), range(n), 0);
        divide(num, elementSum(num), Q);
        maximize(Q, 1e-12);
        // writeMatrix("Q", Q);
        // Compute gradient
        subtract(P, Q, L);
        elementMult(L, num);
        // rowsum = nx1
        DMatrixRMaj rowsum = sumRows(L, null);
        double[] rsum = new double[rowsum.numRows];
        for (int i = 0; i < rsum.length; i++) {
            rsum[i] = rowsum.get(i, 0);
        }
        setDiag(diag, rsum);
        subtract(diag, L, L);
        mult(L, Y, dY);
        scale(4.0, dY);
        // Perform the update
        if (iter < 20)
            momentum = initial_momentum;
        else
            momentum = final_momentum;
        boolean[][] boolMtrx = equal(biggerThan(dY, 0.0), biggerThan(iY, 0.0));
        setData(btNeg, abs(negate(boolMtrx)));
        setData(bt, abs(boolMtrx));
        DMatrixRMaj gainsSmall = new DMatrixRMaj(gains);
        DMatrixRMaj gainsBig = new DMatrixRMaj(gains);
        add(gainsSmall, 0.2);
        scale(0.8, gainsBig);
        elementMult(gainsSmall, btNeg);
        elementMult(gainsBig, bt);
        add(gainsSmall, gainsBig, gains);
        assignAllLessThan(gains, min_gain, min_gain);
        scale(momentum, iY);
        DMatrixRMaj gainsdY = new DMatrixRMaj(gains.numRows, dY.numCols);
        elementMult(gains, dY, gainsdY);
        scale(eta, gainsdY);
        subtractEquals(iY, gainsdY);
        addEquals(Y, iY);
        DMatrixRMaj colMeanY = colMean(Y, 0);
        DMatrixRMaj meanTile = tile(colMeanY, n, 1);
        subtractEquals(Y, meanTile);
        // Compute current value of cost function
        if (iter % 100 == 0) {
            DMatrixRMaj Pdiv = new DMatrixRMaj(P);
            elementDiv(Pdiv, Q);
            elementLog(Pdiv, logdivide);
            replaceNaN(logdivide, Double.MIN_VALUE);
            elementMult(logdivide, P);
            replaceNaN(logdivide, Double.MIN_VALUE);
            double C = elementSum(logdivide);
            end = System.currentTimeMillis();
            monitor.showMessage(TaskMonitor.Level.INFO, String.format("Iteration %d: error is %f (100 iterations in %4.2f seconds)\n", iter, C, (end - start) / 1000.0));
            if (C < 0) {
                monitor.showMessage(TaskMonitor.Level.WARN, "Warning: Error is negative, this is usually a very bad sign!");
            }
            start = System.currentTimeMillis();
        }
        // Stop lying about P-values
        if (iter == 100)
            divide(P, 4);
    }
    end = System.currentTimeMillis();
    monitor.showMessage(TaskMonitor.Level.INFO, String.format("Completed in %4.2f seconds)", (end - total) / 1000.0));
    // Return solution
    return extractDoubleArray(Y);
}

Example 35 with DMatrixRMaj

use of org.ejml.data.DMatrixRMaj in project clusterMaker2 by RBVI.

the class PrincipalComponentAnalysis method response.

/**
 * Computes the dot product of each basis vector against the sample.  Can be used as a measure
 * for membership in the training sample set.  High values correspond to a better fit.
 *
 * @param sample Sample of original data.
 * @return Higher value indicates it is more likely to be a member of input dataset.
 */
public double response(double[] sample) {
    if (sample.length != A.numCols)
        throw new IllegalArgumentException("Expected input vector to be in sample space");
    DMatrixRMaj dots = new DMatrixRMaj(numComponents, 1);
    DMatrixRMaj s = DMatrixRMaj.wrap(A.numCols, 1, sample);
    CommonOps_DDRM.mult(V_t, s, dots);
    return NormOps_DDRM.normF(dots);
}