ILIAS  release_5-4 Revision v5.4.26-12-gabc799a52e6
EigenvalueDecomposition.php
Go to the documentation of this file.
1 <?php
2 
4 
26 {
32  private $n;
33 
39  private $d = [];
40 
41  private $e = [];
42 
48  private $V = [];
49 
55  private $H = [];
56 
62  private $ort;
63 
69  private $cdivr;
70 
71  private $cdivi;
72 
76  private $A;
77 
81  private function tred2(): void
82  {
83  // This is derived from the Algol procedures tred2 by
84  // Bowdler, Martin, Reinsch, and Wilkinson, Handbook for
85  // Auto. Comp., Vol.ii-Linear Algebra, and the corresponding
86  // Fortran subroutine in EISPACK.
87  $this->d = $this->V[$this->n - 1];
88  $j = 0;
89  // Householder reduction to tridiagonal form.
90  for ($i = $this->n - 1; $i > 0; --$i) {
91  $i_ = $i - 1;
92  // Scale to avoid under/overflow.
93  $h = $scale = 0.0;
94  $scale += array_sum(array_map('abs', $this->d));
95  if ($scale == 0.0) {
96  $this->e[$i] = $this->d[$i_];
97  $this->d = array_slice($this->V[$i_], 0, $i_);
98  for ($j = 0; $j < $i; ++$j) {
99  $this->V[$j][$i] = $this->V[$i][$j] = 0.0;
100  }
101  } else {
102  // Generate Householder vector.
103  for ($k = 0; $k < $i; ++$k) {
104  $this->d[$k] /= $scale;
105  $h += $this->d[$k] ** 2;
106  }
107  $f = $this->d[$i_];
108  $g = sqrt($h);
109  if ($f > 0) {
110  $g = -$g;
111  }
112  $this->e[$i] = $scale * $g;
113  $h = $h - $f * $g;
114  $this->d[$i_] = $f - $g;
115  for ($j = 0; $j < $i; ++$j) {
116  $this->e[$j] = 0.0;
117  }
118  // Apply similarity transformation to remaining columns.
119  for ($j = 0; $j < $i; ++$j) {
120  $f = $this->d[$j];
121  $this->V[$j][$i] = $f;
122  $g = $this->e[$j] + $this->V[$j][$j] * $f;
123  for ($k = $j + 1; $k <= $i_; ++$k) {
124  $g += $this->V[$k][$j] * $this->d[$k];
125  $this->e[$k] += $this->V[$k][$j] * $f;
126  }
127  $this->e[$j] = $g;
128  }
129  $f = 0.0;
130  for ($j = 0; $j < $i; ++$j) {
131  $this->e[$j] /= $h;
132  $f += $this->e[$j] * $this->d[$j];
133  }
134  $hh = $f / (2 * $h);
135  for ($j = 0; $j < $i; ++$j) {
136  $this->e[$j] -= $hh * $this->d[$j];
137  }
138  for ($j = 0; $j < $i; ++$j) {
139  $f = $this->d[$j];
140  $g = $this->e[$j];
141  for ($k = $j; $k <= $i_; ++$k) {
142  $this->V[$k][$j] -= ($f * $this->e[$k] + $g * $this->d[$k]);
143  }
144  $this->d[$j] = $this->V[$i - 1][$j];
145  $this->V[$i][$j] = 0.0;
146  }
147  }
148  $this->d[$i] = $h;
149  }
150 
151  // Accumulate transformations.
152  for ($i = 0; $i < $this->n - 1; ++$i) {
153  $this->V[$this->n - 1][$i] = $this->V[$i][$i];
154  $this->V[$i][$i] = 1.0;
155  $h = $this->d[$i + 1];
156  if ($h != 0.0) {
157  for ($k = 0; $k <= $i; ++$k) {
158  $this->d[$k] = $this->V[$k][$i + 1] / $h;
159  }
160  for ($j = 0; $j <= $i; ++$j) {
161  $g = 0.0;
162  for ($k = 0; $k <= $i; ++$k) {
163  $g += $this->V[$k][$i + 1] * $this->V[$k][$j];
164  }
165  for ($k = 0; $k <= $i; ++$k) {
166  $this->V[$k][$j] -= $g * $this->d[$k];
167  }
168  }
169  }
170  for ($k = 0; $k <= $i; ++$k) {
171  $this->V[$k][$i + 1] = 0.0;
172  }
173  }
174 
175  $this->d = $this->V[$this->n - 1];
176  $this->V[$this->n - 1] = array_fill(0, $j, 0.0);
177  $this->V[$this->n - 1][$this->n - 1] = 1.0;
178  $this->e[0] = 0.0;
179  }
180 
189  private function tql2(): void
190  {
191  for ($i = 1; $i < $this->n; ++$i) {
192  $this->e[$i - 1] = $this->e[$i];
193  }
194  $this->e[$this->n - 1] = 0.0;
195  $f = 0.0;
196  $tst1 = 0.0;
197  $eps = 2.0 ** (-52.0);
198 
199  for ($l = 0; $l < $this->n; ++$l) {
200  // Find small subdiagonal element
201  $tst1 = max($tst1, abs($this->d[$l]) + abs($this->e[$l]));
202  $m = $l;
203  while ($m < $this->n) {
204  if (abs($this->e[$m]) <= $eps * $tst1) {
205  break;
206  }
207  ++$m;
208  }
209  // If m == l, $this->d[l] is an eigenvalue,
210  // otherwise, iterate.
211  if ($m > $l) {
212  $iter = 0;
213  do {
214  // Could check iteration count here.
215  ++$iter;
216  // Compute implicit shift
217  $g = $this->d[$l];
218  $p = ($this->d[$l + 1] - $g) / (2.0 * $this->e[$l]);
219  $r = hypo($p, 1.0);
220  if ($p < 0) {
221  $r *= -1;
222  }
223  $this->d[$l] = $this->e[$l] / ($p + $r);
224  $this->d[$l + 1] = $this->e[$l] * ($p + $r);
225  $dl1 = $this->d[$l + 1];
226  $h = $g - $this->d[$l];
227  for ($i = $l + 2; $i < $this->n; ++$i) {
228  $this->d[$i] -= $h;
229  }
230  $f += $h;
231  // Implicit QL transformation.
232  $p = $this->d[$m];
233  $c = 1.0;
234  $c2 = $c3 = $c;
235  $el1 = $this->e[$l + 1];
236  $s = $s2 = 0.0;
237  for ($i = $m - 1; $i >= $l; --$i) {
238  $c3 = $c2;
239  $c2 = $c;
240  $s2 = $s;
241  $g = $c * $this->e[$i];
242  $h = $c * $p;
243  $r = hypo($p, $this->e[$i]);
244  $this->e[$i + 1] = $s * $r;
245  $s = $this->e[$i] / $r;
246  $c = $p / $r;
247  $p = $c * $this->d[$i] - $s * $g;
248  $this->d[$i + 1] = $h + $s * ($c * $g + $s * $this->d[$i]);
249  // Accumulate transformation.
250  for ($k = 0; $k < $this->n; ++$k) {
251  $h = $this->V[$k][$i + 1];
252  $this->V[$k][$i + 1] = $s * $this->V[$k][$i] + $c * $h;
253  $this->V[$k][$i] = $c * $this->V[$k][$i] - $s * $h;
254  }
255  }
256  $p = -$s * $s2 * $c3 * $el1 * $this->e[$l] / $dl1;
257  $this->e[$l] = $s * $p;
258  $this->d[$l] = $c * $p;
259  // Check for convergence.
260  } while (abs($this->e[$l]) > $eps * $tst1);
261  }
262  $this->d[$l] = $this->d[$l] + $f;
263  $this->e[$l] = 0.0;
264  }
265 
266  // Sort eigenvalues and corresponding vectors.
267  for ($i = 0; $i < $this->n - 1; ++$i) {
268  $k = $i;
269  $p = $this->d[$i];
270  for ($j = $i + 1; $j < $this->n; ++$j) {
271  if ($this->d[$j] < $p) {
272  $k = $j;
273  $p = $this->d[$j];
274  }
275  }
276  if ($k != $i) {
277  $this->d[$k] = $this->d[$i];
278  $this->d[$i] = $p;
279  for ($j = 0; $j < $this->n; ++$j) {
280  $p = $this->V[$j][$i];
281  $this->V[$j][$i] = $this->V[$j][$k];
282  $this->V[$j][$k] = $p;
283  }
284  }
285  }
286  }
287 
296  private function orthes(): void
297  {
298  $low = 0;
299  $high = $this->n - 1;
300 
301  for ($m = $low + 1; $m <= $high - 1; ++$m) {
302  // Scale column.
303  $scale = 0.0;
304  for ($i = $m; $i <= $high; ++$i) {
305  $scale = $scale + abs($this->H[$i][$m - 1]);
306  }
307  if ($scale != 0.0) {
308  // Compute Householder transformation.
309  $h = 0.0;
310  for ($i = $high; $i >= $m; --$i) {
311  $this->ort[$i] = $this->H[$i][$m - 1] / $scale;
312  $h += $this->ort[$i] * $this->ort[$i];
313  }
314  $g = sqrt($h);
315  if ($this->ort[$m] > 0) {
316  $g *= -1;
317  }
318  $h -= $this->ort[$m] * $g;
319  $this->ort[$m] -= $g;
320  // Apply Householder similarity transformation
321  // H = (I -u * u' / h) * H * (I -u * u') / h)
322  for ($j = $m; $j < $this->n; ++$j) {
323  $f = 0.0;
324  for ($i = $high; $i >= $m; --$i) {
325  $f += $this->ort[$i] * $this->H[$i][$j];
326  }
327  $f /= $h;
328  for ($i = $m; $i <= $high; ++$i) {
329  $this->H[$i][$j] -= $f * $this->ort[$i];
330  }
331  }
332  for ($i = 0; $i <= $high; ++$i) {
333  $f = 0.0;
334  for ($j = $high; $j >= $m; --$j) {
335  $f += $this->ort[$j] * $this->H[$i][$j];
336  }
337  $f = $f / $h;
338  for ($j = $m; $j <= $high; ++$j) {
339  $this->H[$i][$j] -= $f * $this->ort[$j];
340  }
341  }
342  $this->ort[$m] = $scale * $this->ort[$m];
343  $this->H[$m][$m - 1] = $scale * $g;
344  }
345  }
346 
347  // Accumulate transformations (Algol's ortran).
348  for ($i = 0; $i < $this->n; ++$i) {
349  for ($j = 0; $j < $this->n; ++$j) {
350  $this->V[$i][$j] = ($i == $j ? 1.0 : 0.0);
351  }
352  }
353  for ($m = $high - 1; $m >= $low + 1; --$m) {
354  if ($this->H[$m][$m - 1] != 0.0) {
355  for ($i = $m + 1; $i <= $high; ++$i) {
356  $this->ort[$i] = $this->H[$i][$m - 1];
357  }
358  for ($j = $m; $j <= $high; ++$j) {
359  $g = 0.0;
360  for ($i = $m; $i <= $high; ++$i) {
361  $g += $this->ort[$i] * $this->V[$i][$j];
362  }
363  // Double division avoids possible underflow
364  $g = ($g / $this->ort[$m]) / $this->H[$m][$m - 1];
365  for ($i = $m; $i <= $high; ++$i) {
366  $this->V[$i][$j] += $g * $this->ort[$i];
367  }
368  }
369  }
370  }
371  }
372 
381  private function cdiv($xr, $xi, $yr, $yi): void
382  {
383  if (abs($yr) > abs($yi)) {
384  $r = $yi / $yr;
385  $d = $yr + $r * $yi;
386  $this->cdivr = ($xr + $r * $xi) / $d;
387  $this->cdivi = ($xi - $r * $xr) / $d;
388  } else {
389  $r = $yr / $yi;
390  $d = $yi + $r * $yr;
391  $this->cdivr = ($r * $xr + $xi) / $d;
392  $this->cdivi = ($r * $xi - $xr) / $d;
393  }
394  }
395 
404  private function hqr2(): void
405  {
406  // Initialize
407  $nn = $this->n;
408  $n = $nn - 1;
409  $low = 0;
410  $high = $nn - 1;
411  $eps = 2.0 ** (-52.0);
412  $exshift = 0.0;
413  $p = $q = $r = $s = $z = 0;
414  // Store roots isolated by balanc and compute matrix norm
415  $norm = 0.0;
416 
417  for ($i = 0; $i < $nn; ++$i) {
418  if (($i < $low) || ($i > $high)) {
419  $this->d[$i] = $this->H[$i][$i];
420  $this->e[$i] = 0.0;
421  }
422  for ($j = max($i - 1, 0); $j < $nn; ++$j) {
423  $norm = $norm + abs($this->H[$i][$j]);
424  }
425  }
426 
427  // Outer loop over eigenvalue index
428  $iter = 0;
429  while ($n >= $low) {
430  // Look for single small sub-diagonal element
431  $l = $n;
432  while ($l > $low) {
433  $s = abs($this->H[$l - 1][$l - 1]) + abs($this->H[$l][$l]);
434  if ($s == 0.0) {
435  $s = $norm;
436  }
437  if (abs($this->H[$l][$l - 1]) < $eps * $s) {
438  break;
439  }
440  --$l;
441  }
442  // Check for convergence
443  // One root found
444  if ($l == $n) {
445  $this->H[$n][$n] = $this->H[$n][$n] + $exshift;
446  $this->d[$n] = $this->H[$n][$n];
447  $this->e[$n] = 0.0;
448  --$n;
449  $iter = 0;
450  // Two roots found
451  } elseif ($l == $n - 1) {
452  $w = $this->H[$n][$n - 1] * $this->H[$n - 1][$n];
453  $p = ($this->H[$n - 1][$n - 1] - $this->H[$n][$n]) / 2.0;
454  $q = $p * $p + $w;
455  $z = sqrt(abs($q));
456  $this->H[$n][$n] = $this->H[$n][$n] + $exshift;
457  $this->H[$n - 1][$n - 1] = $this->H[$n - 1][$n - 1] + $exshift;
458  $x = $this->H[$n][$n];
459  // Real pair
460  if ($q >= 0) {
461  if ($p >= 0) {
462  $z = $p + $z;
463  } else {
464  $z = $p - $z;
465  }
466  $this->d[$n - 1] = $x + $z;
467  $this->d[$n] = $this->d[$n - 1];
468  if ($z != 0.0) {
469  $this->d[$n] = $x - $w / $z;
470  }
471  $this->e[$n - 1] = 0.0;
472  $this->e[$n] = 0.0;
473  $x = $this->H[$n][$n - 1];
474  $s = abs($x) + abs($z);
475  $p = $x / $s;
476  $q = $z / $s;
477  $r = sqrt($p * $p + $q * $q);
478  $p = $p / $r;
479  $q = $q / $r;
480  // Row modification
481  for ($j = $n - 1; $j < $nn; ++$j) {
482  $z = $this->H[$n - 1][$j];
483  $this->H[$n - 1][$j] = $q * $z + $p * $this->H[$n][$j];
484  $this->H[$n][$j] = $q * $this->H[$n][$j] - $p * $z;
485  }
486  // Column modification
487  for ($i = 0; $i <= $n; ++$i) {
488  $z = $this->H[$i][$n - 1];
489  $this->H[$i][$n - 1] = $q * $z + $p * $this->H[$i][$n];
490  $this->H[$i][$n] = $q * $this->H[$i][$n] - $p * $z;
491  }
492  // Accumulate transformations
493  for ($i = $low; $i <= $high; ++$i) {
494  $z = $this->V[$i][$n - 1];
495  $this->V[$i][$n - 1] = $q * $z + $p * $this->V[$i][$n];
496  $this->V[$i][$n] = $q * $this->V[$i][$n] - $p * $z;
497  }
498  // Complex pair
499  } else {
500  $this->d[$n - 1] = $x + $p;
501  $this->d[$n] = $x + $p;
502  $this->e[$n - 1] = $z;
503  $this->e[$n] = -$z;
504  }
505  $n = $n - 2;
506  $iter = 0;
507  // No convergence yet
508  } else {
509  // Form shift
510  $x = $this->H[$n][$n];
511  $y = 0.0;
512  $w = 0.0;
513  if ($l < $n) {
514  $y = $this->H[$n - 1][$n - 1];
515  $w = $this->H[$n][$n - 1] * $this->H[$n - 1][$n];
516  }
517  // Wilkinson's original ad hoc shift
518  if ($iter == 10) {
519  $exshift += $x;
520  for ($i = $low; $i <= $n; ++$i) {
521  $this->H[$i][$i] -= $x;
522  }
523  $s = abs($this->H[$n][$n - 1]) + abs($this->H[$n - 1][$n - 2]);
524  $x = $y = 0.75 * $s;
525  $w = -0.4375 * $s * $s;
526  }
527  // MATLAB's new ad hoc shift
528  if ($iter == 30) {
529  $s = ($y - $x) / 2.0;
530  $s = $s * $s + $w;
531  if ($s > 0) {
532  $s = sqrt($s);
533  if ($y < $x) {
534  $s = -$s;
535  }
536  $s = $x - $w / (($y - $x) / 2.0 + $s);
537  for ($i = $low; $i <= $n; ++$i) {
538  $this->H[$i][$i] -= $s;
539  }
540  $exshift += $s;
541  $x = $y = $w = 0.964;
542  }
543  }
544  // Could check iteration count here.
545  $iter = $iter + 1;
546  // Look for two consecutive small sub-diagonal elements
547  $m = $n - 2;
548  while ($m >= $l) {
549  $z = $this->H[$m][$m];
550  $r = $x - $z;
551  $s = $y - $z;
552  $p = ($r * $s - $w) / $this->H[$m + 1][$m] + $this->H[$m][$m + 1];
553  $q = $this->H[$m + 1][$m + 1] - $z - $r - $s;
554  $r = $this->H[$m + 2][$m + 1];
555  $s = abs($p) + abs($q) + abs($r);
556  $p = $p / $s;
557  $q = $q / $s;
558  $r = $r / $s;
559  if ($m == $l) {
560  break;
561  }
562  if (
563  abs($this->H[$m][$m - 1]) * (abs($q) + abs($r)) <
564  $eps * (abs($p) * (abs($this->H[$m - 1][$m - 1]) + abs($z) + abs($this->H[$m + 1][$m + 1])))
565  ) {
566  break;
567  }
568  --$m;
569  }
570  for ($i = $m + 2; $i <= $n; ++$i) {
571  $this->H[$i][$i - 2] = 0.0;
572  if ($i > $m + 2) {
573  $this->H[$i][$i - 3] = 0.0;
574  }
575  }
576  // Double QR step involving rows l:n and columns m:n
577  for ($k = $m; $k <= $n - 1; ++$k) {
578  $notlast = ($k != $n - 1);
579  if ($k != $m) {
580  $p = $this->H[$k][$k - 1];
581  $q = $this->H[$k + 1][$k - 1];
582  $r = ($notlast ? $this->H[$k + 2][$k - 1] : 0.0);
583  $x = abs($p) + abs($q) + abs($r);
584  if ($x != 0.0) {
585  $p = $p / $x;
586  $q = $q / $x;
587  $r = $r / $x;
588  }
589  }
590  if ($x == 0.0) {
591  break;
592  }
593  $s = sqrt($p * $p + $q * $q + $r * $r);
594  if ($p < 0) {
595  $s = -$s;
596  }
597  if ($s != 0) {
598  if ($k != $m) {
599  $this->H[$k][$k - 1] = -$s * $x;
600  } elseif ($l != $m) {
601  $this->H[$k][$k - 1] = -$this->H[$k][$k - 1];
602  }
603  $p = $p + $s;
604  $x = $p / $s;
605  $y = $q / $s;
606  $z = $r / $s;
607  $q = $q / $p;
608  $r = $r / $p;
609  // Row modification
610  for ($j = $k; $j < $nn; ++$j) {
611  $p = $this->H[$k][$j] + $q * $this->H[$k + 1][$j];
612  if ($notlast) {
613  $p = $p + $r * $this->H[$k + 2][$j];
614  $this->H[$k + 2][$j] = $this->H[$k + 2][$j] - $p * $z;
615  }
616  $this->H[$k][$j] = $this->H[$k][$j] - $p * $x;
617  $this->H[$k + 1][$j] = $this->H[$k + 1][$j] - $p * $y;
618  }
619  // Column modification
620  $iMax = min($n, $k + 3);
621  for ($i = 0; $i <= $iMax; ++$i) {
622  $p = $x * $this->H[$i][$k] + $y * $this->H[$i][$k + 1];
623  if ($notlast) {
624  $p = $p + $z * $this->H[$i][$k + 2];
625  $this->H[$i][$k + 2] = $this->H[$i][$k + 2] - $p * $r;
626  }
627  $this->H[$i][$k] = $this->H[$i][$k] - $p;
628  $this->H[$i][$k + 1] = $this->H[$i][$k + 1] - $p * $q;
629  }
630  // Accumulate transformations
631  for ($i = $low; $i <= $high; ++$i) {
632  $p = $x * $this->V[$i][$k] + $y * $this->V[$i][$k + 1];
633  if ($notlast) {
634  $p = $p + $z * $this->V[$i][$k + 2];
635  $this->V[$i][$k + 2] = $this->V[$i][$k + 2] - $p * $r;
636  }
637  $this->V[$i][$k] = $this->V[$i][$k] - $p;
638  $this->V[$i][$k + 1] = $this->V[$i][$k + 1] - $p * $q;
639  }
640  } // ($s != 0)
641  } // k loop
642  } // check convergence
643  } // while ($n >= $low)
644 
645  // Backsubstitute to find vectors of upper triangular form
646  if ($norm == 0.0) {
647  return;
648  }
649 
650  for ($n = $nn - 1; $n >= 0; --$n) {
651  $p = $this->d[$n];
652  $q = $this->e[$n];
653  // Real vector
654  if ($q == 0) {
655  $l = $n;
656  $this->H[$n][$n] = 1.0;
657  for ($i = $n - 1; $i >= 0; --$i) {
658  $w = $this->H[$i][$i] - $p;
659  $r = 0.0;
660  for ($j = $l; $j <= $n; ++$j) {
661  $r = $r + $this->H[$i][$j] * $this->H[$j][$n];
662  }
663  if ($this->e[$i] < 0.0) {
664  $z = $w;
665  $s = $r;
666  } else {
667  $l = $i;
668  if ($this->e[$i] == 0.0) {
669  if ($w != 0.0) {
670  $this->H[$i][$n] = -$r / $w;
671  } else {
672  $this->H[$i][$n] = -$r / ($eps * $norm);
673  }
674  // Solve real equations
675  } else {
676  $x = $this->H[$i][$i + 1];
677  $y = $this->H[$i + 1][$i];
678  $q = ($this->d[$i] - $p) * ($this->d[$i] - $p) + $this->e[$i] * $this->e[$i];
679  $t = ($x * $s - $z * $r) / $q;
680  $this->H[$i][$n] = $t;
681  if (abs($x) > abs($z)) {
682  $this->H[$i + 1][$n] = (-$r - $w * $t) / $x;
683  } else {
684  $this->H[$i + 1][$n] = (-$s - $y * $t) / $z;
685  }
686  }
687  // Overflow control
688  $t = abs($this->H[$i][$n]);
689  if (($eps * $t) * $t > 1) {
690  for ($j = $i; $j <= $n; ++$j) {
691  $this->H[$j][$n] = $this->H[$j][$n] / $t;
692  }
693  }
694  }
695  }
696  // Complex vector
697  } elseif ($q < 0) {
698  $l = $n - 1;
699  // Last vector component imaginary so matrix is triangular
700  if (abs($this->H[$n][$n - 1]) > abs($this->H[$n - 1][$n])) {
701  $this->H[$n - 1][$n - 1] = $q / $this->H[$n][$n - 1];
702  $this->H[$n - 1][$n] = -($this->H[$n][$n] - $p) / $this->H[$n][$n - 1];
703  } else {
704  $this->cdiv(0.0, -$this->H[$n - 1][$n], $this->H[$n - 1][$n - 1] - $p, $q);
705  $this->H[$n - 1][$n - 1] = $this->cdivr;
706  $this->H[$n - 1][$n] = $this->cdivi;
707  }
708  $this->H[$n][$n - 1] = 0.0;
709  $this->H[$n][$n] = 1.0;
710  for ($i = $n - 2; $i >= 0; --$i) {
711  // double ra,sa,vr,vi;
712  $ra = 0.0;
713  $sa = 0.0;
714  for ($j = $l; $j <= $n; ++$j) {
715  $ra = $ra + $this->H[$i][$j] * $this->H[$j][$n - 1];
716  $sa = $sa + $this->H[$i][$j] * $this->H[$j][$n];
717  }
718  $w = $this->H[$i][$i] - $p;
719  if ($this->e[$i] < 0.0) {
720  $z = $w;
721  $r = $ra;
722  $s = $sa;
723  } else {
724  $l = $i;
725  if ($this->e[$i] == 0) {
726  $this->cdiv(-$ra, -$sa, $w, $q);
727  $this->H[$i][$n - 1] = $this->cdivr;
728  $this->H[$i][$n] = $this->cdivi;
729  } else {
730  // Solve complex equations
731  $x = $this->H[$i][$i + 1];
732  $y = $this->H[$i + 1][$i];
733  $vr = ($this->d[$i] - $p) * ($this->d[$i] - $p) + $this->e[$i] * $this->e[$i] - $q * $q;
734  $vi = ($this->d[$i] - $p) * 2.0 * $q;
735  if ($vr == 0.0 & $vi == 0.0) {
736  $vr = $eps * $norm * (abs($w) + abs($q) + abs($x) + abs($y) + abs($z));
737  }
738  $this->cdiv($x * $r - $z * $ra + $q * $sa, $x * $s - $z * $sa - $q * $ra, $vr, $vi);
739  $this->H[$i][$n - 1] = $this->cdivr;
740  $this->H[$i][$n] = $this->cdivi;
741  if (abs($x) > (abs($z) + abs($q))) {
742  $this->H[$i + 1][$n - 1] = (-$ra - $w * $this->H[$i][$n - 1] + $q * $this->H[$i][$n]) / $x;
743  $this->H[$i + 1][$n] = (-$sa - $w * $this->H[$i][$n] - $q * $this->H[$i][$n - 1]) / $x;
744  } else {
745  $this->cdiv(-$r - $y * $this->H[$i][$n - 1], -$s - $y * $this->H[$i][$n], $z, $q);
746  $this->H[$i + 1][$n - 1] = $this->cdivr;
747  $this->H[$i + 1][$n] = $this->cdivi;
748  }
749  }
750  // Overflow control
751  $t = max(abs($this->H[$i][$n - 1]), abs($this->H[$i][$n]));
752  if (($eps * $t) * $t > 1) {
753  for ($j = $i; $j <= $n; ++$j) {
754  $this->H[$j][$n - 1] = $this->H[$j][$n - 1] / $t;
755  $this->H[$j][$n] = $this->H[$j][$n] / $t;
756  }
757  }
758  } // end else
759  } // end for
760  } // end else for complex case
761  } // end for
762 
763  // Vectors of isolated roots
764  for ($i = 0; $i < $nn; ++$i) {
765  if ($i < $low | $i > $high) {
766  for ($j = $i; $j < $nn; ++$j) {
767  $this->V[$i][$j] = $this->H[$i][$j];
768  }
769  }
770  }
771 
772  // Back transformation to get eigenvectors of original matrix
773  for ($j = $nn - 1; $j >= $low; --$j) {
774  for ($i = $low; $i <= $high; ++$i) {
775  $z = 0.0;
776  $kMax = min($j, $high);
777  for ($k = $low; $k <= $kMax; ++$k) {
778  $z = $z + $this->V[$i][$k] * $this->H[$k][$j];
779  }
780  $this->V[$i][$j] = $z;
781  }
782  }
783  }
784 
785  // end hqr2
786 
792  public function __construct(Matrix $Arg)
793  {
794  $this->A = $Arg->getArray();
795  $this->n = $Arg->getColumnDimension();
796 
797  $issymmetric = true;
798  for ($j = 0; ($j < $this->n) & $issymmetric; ++$j) {
799  for ($i = 0; ($i < $this->n) & $issymmetric; ++$i) {
800  $issymmetric = ($this->A[$i][$j] == $this->A[$j][$i]);
801  }
802  }
803 
804  if ($issymmetric) {
805  $this->V = $this->A;
806  // Tridiagonalize.
807  $this->tred2();
808  // Diagonalize.
809  $this->tql2();
810  } else {
811  $this->H = $this->A;
812  $this->ort = [];
813  // Reduce to Hessenberg form.
814  $this->orthes();
815  // Reduce Hessenberg to real Schur form.
816  $this->hqr2();
817  }
818  }
819 
825  public function getV()
826  {
827  return new Matrix($this->V, $this->n, $this->n);
828  }
829 
835  public function getRealEigenvalues()
836  {
837  return $this->d;
838  }
839 
845  public function getImagEigenvalues()
846  {
847  return $this->e;
848  }
849 
855  public function getD()
856  {
857  $D = [];
858  for ($i = 0; $i < $this->n; ++$i) {
859  $D[$i] = array_fill(0, $this->n, 0.0);
860  $D[$i][$i] = $this->d[$i];
861  if ($this->e[$i] == 0) {
862  continue;
863  }
864  $o = ($this->e[$i] > 0) ? $i + 1 : $i - 1;
865  $D[$i][$o] = $this->e[$i];
866  }
867 
868  return new Matrix($D);
869  }
870 }
Class to obtain eigenvalues and eigenvectors of a real matrix.
getRealEigenvalues()
Return the real parts of the eigenvalues.
getImagEigenvalues()
Return the imaginary parts of the eigenvalues.
$h
if(! $in) print Initializing normalization quick check tables n
$s
Definition: pwgen.php:45
$w
$r
Definition: example_031.php:79
$y
Definition: example_007.php:83
$eps
Definition: metadata.php:61
hypo($a, $b)
Pythagorean Theorem:.
Definition: Maths.php:18
global $l
Definition: afr.php:30
getColumnDimension()
getColumnDimension.
Definition: Matrix.php:137
$i
Definition: disco.tpl.php:19
__construct(Matrix $Arg)
Constructor: Check for symmetry, then construct the eigenvalue decomposition.
hqr2()
Nonsymmetric reduction from Hessenberg to real Schur form.
Class for the creating "special" Matrices.
Definition: Builder.php:11
$x
Definition: complexTest.php:9
tred2()
Symmetric Householder reduction to tridiagonal form.