| Source for file SingularValueDecomposition.phpDocumentation is available at SingularValueDecomposition.php 
 *     For an m-by-n matrix A with m >= n, the singular value decomposition is *     an m-by-n orthogonal matrix U, an n-by-n diagonal matrix S, and *     an n-by-n orthogonal matrix V so that A = U*S*V'. *     The singular values, sigma[$k] = S[$k][$k], are ordered so that *     sigma[0] >= sigma[1] >= ... >= sigma[n-1]. *     The singular value decompostion always exists, so the constructor will *     never fail.  The matrix condition number and the effective numerical *     rank can be computed from this decomposition.     *    Internal storage of singular values.     *    Construct the singular value decomposition     *    Derived from LINPACK code.     *    @param $A Rectangular matrix     *    @return Structure to access U, S and V.        $A = $Arg->getArrayCopy();        $this->m = $Arg->getRowDimension();        $this->n = $Arg->getColumnDimension();        $nu      = min($this->m, $this->n);        $nct = min($this->m - 1, $this->n);        $nrt = max(0, min($this->n - 2, $this->m));        // Reduce A to bidiagonal form, storing the diagonal elements        // in s and the super-diagonal elements in e.        for ($k = 0; $k < max($nct,$nrt); ++$k) {                // Compute the transformation for the k-th column and                // place the k-th diagonal in s[$k].                // Compute 2-norm of k-th column without under/overflow.                for ($i = $k; $i < $this->m; ++$i) {                    $this->s[$k] = hypo($this->s[$k], $A[$i][$k]);                if ($this->s[$k] != 0.0) {                        $this->s[$k] = -$this->s[$k];                    for ($i = $k; $i < $this->m; ++$i) {                        $A[$i][$k] /= $this->s[$k];                $this->s[$k] = -$this->s[$k];            for ($j = $k + 1; $j < $this->n; ++$j) {                if (($k < $nct) & ($this->s[$k] != 0.0)) {                    // Apply the transformation.                    for ($i = $k; $i < $this->m; ++$i) {                        $t += $A[$i][$k] * $A[$i][$j];                    for ($i = $k; $i < $this->m; ++$i) {                        $A[$i][$j] += $t * $A[$i][$k];                    // Place the k-th row of A into e for the                    // subsequent calculation of the row transformation.            if ($wantu AND ($k < $nct)) {                // Place the transformation in U for subsequent back                for ($i = $k; $i < $this->m; ++$i) {                    $this->U[$i][$k] = $A[$i][$k];                // Compute the k-th row transformation and place the                // k-th super-diagonal in e[$k].                // Compute 2-norm without under/overflow.                for ($i = $k + 1; $i < $this->n; ++$i) {                    $e[$k] = hypo($e[$k], $e[$i]);                    for ($i = $k + 1; $i < $this->n; ++$i) {                if (($k+1 < $this->m) AND ($e[$k] != 0.0)) {                    // Apply the transformation.                    for ($i = $k +1; $i < $this->m; ++$i) {                    for ($j = $k +1; $j < $this->n; ++$j) {                        for ($i = $k +1; $i < $this->m; ++$i) {                            $work[$i] += $e[$j] * $A[$i][$j];                    for ($j = $k + 1; $j < $this->n; ++$j) {                        for ($i = $k + 1; $i < $this->m; ++$i) {                            $A[$i][$j] += $t * $work[$i];                    // Place the transformation in V for subsequent                    for ($i = $k + 1; $i < $this->n; ++$i) {                        $this->V[$i][$k] = $e[$i];        // Set up the final bidiagonal matrix or order p.        $p = min($this->n, $this->m + 1);            $this->s[$nct] = $A[$nct][$nct];            $e[$nrt] = $A[$nrt][$p -1];        // If required, generate U.            for ($j = $nct; $j < $nu; ++$j) {                for ($i = 0; $i < $this->m; ++$i) {            for ($k = $nct - 1; $k >= 0; --$k) {                if ($this->s[$k] != 0.0) {                    for ($j = $k + 1; $j < $nu; ++$j) {                        for ($i = $k; $i < $this->m; ++$i) {                            $t += $this->U[$i][$k] * $this->U[$i][$j];                        $t = -$t / $this->U[$k][$k];                        for ($i = $k; $i < $this->m; ++$i) {                            $this->U[$i][$j] += $t * $this->U[$i][$k];                    for ($i = $k; $i < $this->m; ++$i ) {                        $this->U[$i][$k] = -$this->U[$i][$k];                    $this->U[$k][$k] = 1.0 + $this->U[$k][$k];                    for ($i = 0; $i < $k - 1; ++$i) {                    for ($i = 0; $i < $this->m; ++$i) {        // If required, generate V.            for ($k = $this->n - 1; $k >= 0; --$k) {                if (($k < $nrt) AND ($e[$k] != 0.0)) {                    for ($j = $k + 1; $j < $nu; ++$j) {                        for ($i = $k + 1; $i < $this->n; ++$i) {                            $t += $this->V[$i][$k] * $this->V[$i][$j];                        $t = -$t / $this->V[$k +1][$k];                        for ($i = $k + 1; $i < $this->n; ++$i) {                            $this->V[$i][$j] += $t * $this->V[$i][$k];                for ($i = 0; $i < $this->n; ++$i) {        // Main iteration loop for the singular values.            // Here is where a test for too many iterations would go.            // This section of the program inspects for negligible            // elements in the s and e arrays.  On completion the            // variables kase and k are set as follows:            // kase = 1  if s(p) and e[k-1] are negligible and k<p            // kase = 2  if s(k) is negligible and k<p            // kase = 3  if e[k-1] is negligible, k<p, and            //           s(k), ..., s(p) are not negligible (qr step).            // kase = 4  if e(p-1) is negligible (convergence).            for ($k = $p - 2; $k >= -1; --$k) {                if (abs($e[$k]) <= $eps * (abs($this->s[$k]) + abs($this->s[$k +1]))) {                for ($ks = $p - 1; $ks >= $k; --$ks) {                    $t = ($ks != $p ? abs($e[$ks]) : 0.) + ($ks != $k + 1 ? abs($e[$ks -1]) : 0.);                    if (abs($this->s[$ks]) <= $eps * $t)  {                } else if ($ks == $p -1) {            // Perform the task indicated by kase.                // Deflate negligible s(p).                        for ($j = $p - 2; $j >= $k; --$j) {                            $t  = hypo($this->s[$j],$f);                                $e[$j-1] = $cs * $e[$j -1];                                for ($i = 0; $i < $this->n; ++$i) {                                    $t = $cs * $this->V[$i][$j] + $sn * $this->V[$i][$p -1];                                    $this->V[$i][$p-1] = -$sn * $this->V[$i][$j] + $cs * $this->V[$i][$p -1];                // Split at negligible s(k).                        for ($j = $k; $j < $p; ++$j) {                            $t = hypo($this->s[$j], $f);                                for ($i = 0; $i < $this->m; ++$i) {                                    $t = $cs * $this->U[$i][$j] + $sn * $this->U[$i][$k -1];                                    $this->U[$i][$k-1] = -$sn * $this->U[$i][$j] + $cs * $this->U[$i][$k -1];                                    abs($this->s[$p-1]),abs($this->s[$p -2])),abs($e[$p -2])),                                    abs($this->s[$k])), abs($e[$k]));                        $sp   = $this->s[$p -1] / $scale;                        $spm1 = $this->s[$p -2] / $scale;                        $epm1 = $e[$p -2] / $scale;                        $sk   = $this->s[$k] / $scale;                        $b    = (($spm1 + $sp) * ($spm1 - $sp) + $epm1 * $epm1) / 2.0;                        $c    = ($sp * $epm1) * ($sp * $epm1);                        if (($b != 0.0) || ($c != 0.0)) {                            $shift = sqrt($b * $b + $c);                            $shift = $c / ($b + $shift);                        $f = ($sk + $sp) * ($sk - $sp) + $shift;                        for ($j = $k; $j < $p -1; ++$j) {                            $f = $cs * $this->s[$j] + $sn * $e[$j];                            $e[$j] = $cs * $e[$j] - $sn * $this->s[$j];                            $g = $sn * $this->s[$j +1];                            $this->s[$j+1] = $cs * $this->s[$j +1];                                for ($i = 0; $i < $this->n; ++$i) {                                    $t = $cs * $this->V[$i][$j] + $sn * $this->V[$i][$j +1];                                    $this->V[$i][$j+1] = -$sn * $this->V[$i][$j] + $cs * $this->V[$i][$j +1];                            $f = $cs * $e[$j] + $sn * $this->s[$j +1];                            $this->s[$j+1] = -$sn * $e[$j] + $cs * $this->s[$j +1];                            $e[$j+1] = $cs * $e[$j +1];                            if ($wantu && ($j < $this->m - 1)) {                                for ($i = 0; $i < $this->m; ++$i) {                                    $t = $cs * $this->U[$i][$j] + $sn * $this->U[$i][$j +1];                                    $this->U[$i][$j+1] = -$sn * $this->U[$i][$j] + $cs * $this->U[$i][$j +1];                        // Make the singular values positive.                        if ($this->s[$k] <= 0.0) {                            $this->s[$k] = ($this->s[$k] < 0.0 ? -$this->s[$k] : 0.0);                                for ($i = 0; $i <= $pp; ++$i) {                                    $this->V[$i][$k] = -$this->V[$i][$k];                        // Order the singular values.                            if ($this->s[$k] >= $this->s[$k +1]) {                            $this->s[$k] = $this->s[$k +1];                            if ($wantv AND ($k < $this->n - 1)) {                                for ($i = 0; $i < $this->n; ++$i) {                                    $this->V[$i][$k+1] = $this->V[$i][$k];                            if ($wantu AND ($k < $this->m -1)) {                                for ($i = 0; $i < $this->m; ++$i) {                                    $this->U[$i][$k+1] = $this->U[$i][$k];     *    Return the left singular vectors        return new Matrix($this->U, $this->m, min($this->m + 1, $this->n));     *    Return the right singular vectors        return new Matrix($this->V);     *    Return the one-dimensional array of singular values     *    Return the diagonal matrix of singular values        for ($i = 0; $i < $this->n; ++$i) {            for ($j = 0; $j < $this->n; ++$j) {            $S[$i][$i] = $this->s[$i];    public function norm2() {     *    Two norm condition number        return $this->s[0] / $this->s[min($this->m, $this->n) - 1];     *    Effective numerical matrix rank     *    @return Number of nonnegligible singular values.        $tol = max($this->m, $this->n) * $this->s[0] * $eps;        for ($i = 0; $i < count($this->s); ++$i) {            if ($this->s[$i] > $tol) {}    //    class SingularValueDecomposition |