1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
declare(strict_types=1); |
4
|
|
|
|
5
|
|
|
namespace Np; |
6
|
|
|
|
7
|
|
|
use Np\core\nd; |
8
|
|
|
use Np\core\blas; |
9
|
|
|
use Np\core\lapack; |
10
|
|
|
use Np\reductions\ref; |
11
|
|
|
use Np\reductions\rref; |
12
|
|
|
use Np\decompositions\lu; |
13
|
|
|
use Np\decompositions\svd; |
14
|
|
|
use Np\decompositions\eigen; |
15
|
|
|
use Np\decompositions\cholesky; |
16
|
|
|
|
17
|
|
|
/** |
18
|
|
|
* Matrix |
19
|
|
|
* A fast lite memory efficient Scientific Computing for php |
20
|
|
|
* |
21
|
|
|
* @package NumPhp |
22
|
|
|
* @category Scientific Computing |
23
|
|
|
* @author ghost (Shubham Chaudhary) |
24
|
|
|
* @email [email protected] |
25
|
|
|
* @copyright (c) 2020-2021, Shubham Chaudhary |
26
|
|
|
*/ |
27
|
|
|
class matrix extends nd{ |
28
|
|
|
|
29
|
|
|
use ops,linAlg; |
30
|
|
|
/** |
31
|
|
|
* create empty 2d matrix for given data type |
32
|
|
|
* @param int $row num of rows |
33
|
|
|
* @param int $col num of cols |
34
|
|
|
* @param int $dtype matrix data type float|double |
35
|
|
|
* @return \Np\matrix |
36
|
|
|
*/ |
37
|
|
|
public static function factory(int $row, int $col, int $dtype = self::FLOAT): matrix { |
38
|
|
|
return new self($row, $col, $dtype); |
39
|
|
|
} |
40
|
|
|
|
41
|
|
|
/** |
42
|
|
|
* create 2d matrix using php array |
43
|
|
|
* @param array $data |
44
|
|
|
* @param int $dtype matrix data type float|double |
45
|
|
|
* @return \Np\matrix |
46
|
|
|
*/ |
47
|
|
|
public static function ar(array $data, int $dtype = self::FLOAT): matrix { |
48
|
|
|
if (is_array($data) && is_array($data[0])) { |
49
|
|
|
$ar = self::factory(count($data), count($data[0]), $dtype); |
50
|
|
|
$ar->setData($data); |
51
|
|
|
unset($data); |
52
|
|
|
return $ar; |
53
|
|
|
} else { |
54
|
|
|
self::_err('given array is not rank-2 or given is not an array'); |
55
|
|
|
} |
56
|
|
|
} |
57
|
|
|
|
58
|
|
|
/** |
59
|
|
|
* create one like 2d matrix |
60
|
|
|
* @param int $row |
61
|
|
|
* @param int $col |
62
|
|
|
* @return \Np\matrix |
63
|
|
|
*/ |
64
|
|
|
public static function ones(int $row, int $col, int $dtype = self::FLOAT): matrix { |
65
|
|
|
$ar = self::factory($row, $col, $dtype); |
66
|
|
|
for ($i = 0; $i < $ar->ndim; ++$i) { |
67
|
|
|
$ar->data[$i] = 1; |
68
|
|
|
} |
69
|
|
|
return $ar; |
70
|
|
|
} |
71
|
|
|
|
72
|
|
|
/** |
73
|
|
|
* Create Matrix with random values |
74
|
|
|
* @param int $row |
75
|
|
|
* @param int $col |
76
|
|
|
* @param int $dtype Float|Double |
77
|
|
|
* @return \Np\matrix |
78
|
|
|
*/ |
79
|
|
|
public static function randn(int $row, int $col, int $dtype = self::FLOAT): matrix { |
80
|
|
|
$ar = self::factory($row, $col, $dtype); |
81
|
|
|
$max = getrandmax(); |
82
|
|
|
for ($i = 0; $i < $ar->ndim; ++$i) { |
83
|
|
|
$ar->data[$i] = rand() / $max; |
84
|
|
|
} |
85
|
|
|
return $ar; |
86
|
|
|
} |
87
|
|
|
|
88
|
|
|
/** |
89
|
|
|
* Return 2d matrix with uniform values |
90
|
|
|
* @param int $row |
91
|
|
|
* @param int $col |
92
|
|
|
* @param int $dtype |
93
|
|
|
* @return \Np\matrix |
94
|
|
|
*/ |
95
|
|
|
public static function uniform(int $row, int $col, int $dtype = self::FLOAT): matrix { |
96
|
|
|
$ar = self::factory($row, $col, $dtype); |
97
|
|
|
$max = getrandmax(); |
98
|
|
|
for ($i = 0; $i < $ar->ndim; ++$i) { |
99
|
|
|
$ar->data[$i] = rand(-$max, $max) / $max; |
100
|
|
|
} |
101
|
|
|
return $ar; |
102
|
|
|
} |
103
|
|
|
|
104
|
|
|
/** |
105
|
|
|
* Return a zero matrix with the given dimensions. |
106
|
|
|
* @param int $row |
107
|
|
|
* @param int $col |
108
|
|
|
* @param int $dtype |
109
|
|
|
* @return \Np\matrix |
110
|
|
|
*/ |
111
|
|
|
public static function zeros(int $row, int $col, int $dtype = self::FLOAT): matrix { |
112
|
|
|
$ar = self::factory($row, $col, $dtype); |
113
|
|
|
for ($i = 0; $i < $ar->ndim; ++$i) { |
114
|
|
|
$ar->data[$i] = 0.0; |
115
|
|
|
} |
116
|
|
|
return $ar; |
117
|
|
|
} |
118
|
|
|
|
119
|
|
|
/** |
120
|
|
|
* create a null like 2d matrix |
121
|
|
|
* @param int $row |
122
|
|
|
* @param int $col |
123
|
|
|
* @return \Np\matrix |
124
|
|
|
*/ |
125
|
|
|
public static function null(int $row, int $col, int $dtype = self::FLOAT): matrix { |
126
|
|
|
$ar = self::factory($row, $col, $dtype); |
127
|
|
|
for ($i = 0; $i < $ar->ndim; ++$i) { |
128
|
|
|
$ar->data[$i] = null; |
129
|
|
|
} |
130
|
|
|
return $ar; |
131
|
|
|
} |
132
|
|
|
|
133
|
|
|
/** |
134
|
|
|
* create a 2d matrix with given scalar value |
135
|
|
|
* @param int $row |
136
|
|
|
* @param int $col |
137
|
|
|
* @param int|float|double $val |
138
|
|
|
* @return \Np\matrix |
139
|
|
|
*/ |
140
|
|
|
public static function full(int $row, int $col, $val, int $dtype = self::FLOAT): matrix { |
141
|
|
|
$ar = self::factory($row, $col, $dtype); |
142
|
|
|
for ($i = 0; $i < $ar->ndim; ++$i) { |
143
|
|
|
$ar->data[$i] = $val; |
144
|
|
|
} |
145
|
|
|
return $ar; |
146
|
|
|
} |
147
|
|
|
|
148
|
|
|
/** |
149
|
|
|
* create a diagonal 2d matrix with given 1d array; |
150
|
|
|
* @param array $elements |
151
|
|
|
* @return \Np\matrix |
152
|
|
|
*/ |
153
|
|
|
public static function diagonal(array $elements, int $dtype = self::FLOAT): matrix { |
154
|
|
|
$n = count($elements); |
155
|
|
|
$ar = self::factory($n, $n, $dtype); |
156
|
|
|
for ($i = 0; $i < $n; ++$i) { |
157
|
|
|
$ar->data[$i * $n + $i] = $elements[$i]; #for ($j = 0; $j < $n; ++$j) {$i === $j ? $elements[$i] : 0;#} |
158
|
|
|
} |
159
|
|
|
return $ar; |
160
|
|
|
} |
161
|
|
|
|
162
|
|
|
/** |
163
|
|
|
* Generate a m x n matrix with elements from a Poisson distribution. |
164
|
|
|
* @param int $row |
165
|
|
|
* @param int $col |
166
|
|
|
* @param float $lambda |
167
|
|
|
* @param int $dtype |
168
|
|
|
* @return \Np\matrix |
169
|
|
|
*/ |
170
|
|
|
public static function poisson(int $row, int $col, float $lambda = 1.0, int $dtype = self::FLOAT): matrix { |
171
|
|
|
$ar = self::factory($row, $col, $dtype); |
172
|
|
|
$max = getrandmax(); |
173
|
|
|
$l = exp(-$lambda); |
174
|
|
|
for ($i = 0; $i < $row; ++$i) { |
175
|
|
|
for ($j = 0; $j < $col; ++$j) { |
176
|
|
|
$k = 0; |
177
|
|
|
$p = 1.0; |
178
|
|
|
while ($p > $l) { |
179
|
|
|
++$k; |
180
|
|
|
$p = $p * rand() / $max; |
181
|
|
|
} |
182
|
|
|
$ar->data[$i * $col + $j] = $k - 1; |
183
|
|
|
} |
184
|
|
|
} |
185
|
|
|
return $ar; |
186
|
|
|
} |
187
|
|
|
|
188
|
|
|
/** |
189
|
|
|
* Return a standard normally distributed random matrix i.e values |
190
|
|
|
* between -1 and 1. |
191
|
|
|
* @param int $row |
192
|
|
|
* @param int $col |
193
|
|
|
* @param int $dtype Description |
194
|
|
|
* @return \Np\matrix |
195
|
|
|
*/ |
196
|
|
|
public static function gaussian(int $row, int $col, int $dtype = self::FLOAT): matrix { |
197
|
|
|
$max = getrandmax(); |
198
|
|
|
$a = $extras = []; |
199
|
|
|
|
200
|
|
|
while (count($a) < $row) { |
201
|
|
|
$rowA = []; |
202
|
|
|
|
203
|
|
|
if (!empty($extras)) { |
204
|
|
|
$rowA[] = array_pop($extras); |
205
|
|
|
} |
206
|
|
|
|
207
|
|
|
while (count($rowA) < $col) { |
208
|
|
|
$r = sqrt(-2.0 * log(rand() / $max)); |
209
|
|
|
|
210
|
|
|
$phi = rand() / $max * self::TWO_PI; |
211
|
|
|
|
212
|
|
|
$rowA[] = $r * sin($phi); |
213
|
|
|
$rowA[] = $r * cos($phi); |
214
|
|
|
} |
215
|
|
|
|
216
|
|
|
if (count($rowA) > $col) { |
217
|
|
|
$extras[] = array_pop($rowA); |
218
|
|
|
} |
219
|
|
|
|
220
|
|
|
$a[] = $rowA; |
221
|
|
|
} |
222
|
|
|
|
223
|
|
|
return self::ar($a, $dtype); |
224
|
|
|
} |
225
|
|
|
|
226
|
|
|
/** |
227
|
|
|
* create an identity matrix with the given dimensions. |
228
|
|
|
* @param int $n |
229
|
|
|
* @param int $dtype |
230
|
|
|
* @return matrix |
231
|
|
|
* @throws \InvalidArgumentException |
232
|
|
|
*/ |
233
|
|
|
public static function identity(int $n, int $dtype = self::FLOAT): matrix { |
234
|
|
|
if ($n < 1) { |
235
|
|
|
self::_dimensionaMisMatchErr('dimensionality must be greater than 0 on all axes.'); |
236
|
|
|
} |
237
|
|
|
|
238
|
|
|
$ar = self::factory($n, $n, $dtype); |
239
|
|
|
for ($i = 0; $i < $n; ++$i) { |
240
|
|
|
for ($j = 0; $j < $n; ++$j) { |
241
|
|
|
$ar->data[$i * $n + $j] = $i === $j ? 1 : 0; |
242
|
|
|
} |
243
|
|
|
} |
244
|
|
|
return $ar; |
245
|
|
|
} |
246
|
|
|
|
247
|
|
|
/** |
248
|
|
|
* 2D convolution between a matrix ma and kernel kb, with a given stride. |
249
|
|
|
* @param \Np\matrix $m |
250
|
|
|
* @param int $stride |
251
|
|
|
* @return matrix |
252
|
|
|
*/ |
253
|
|
|
public function convolve(matrix $m, int $stride = 1): matrix { |
254
|
|
|
return convolve::conv2D($this, $m, $stride); |
255
|
|
|
} |
256
|
|
|
|
257
|
|
|
/** |
258
|
|
|
* Calculate the determinant of the matrix. |
259
|
|
|
* @return float |
260
|
|
|
*/ |
261
|
|
|
public function det(): float { |
262
|
|
|
if (!$this->isSquare()) { |
263
|
|
|
self::_err('determinant is undefined for a non square matrix'); |
264
|
|
|
} |
265
|
|
|
$lu = $this->lu(); |
266
|
|
|
$nSwaps = $lu->p()->diagonalAsVector()->subtract($lu->p()->diagonalAsVector()->sum())->col - 1; |
267
|
|
|
$detP = (-1) ** $nSwaps; |
268
|
|
|
$detL = $lu->l()->diagonalAsVector()->product(); |
269
|
|
|
$detU = $lu->u()->diagonalAsVector()->product(); |
270
|
|
|
unset($lu); |
271
|
|
|
return ($detP * $detL * $detU); |
272
|
|
|
} |
273
|
|
|
|
274
|
|
|
/** |
275
|
|
|
* Return the trace of the matrix i.e the sum of all diagonal elements of a square matrix. |
276
|
|
|
* @return float |
277
|
|
|
*/ |
278
|
|
|
public function trace(): float { |
279
|
|
|
if (!$this->isSquare()) { |
280
|
|
|
self::_err('Error::matrix is not a squared can not Trace!'); |
281
|
|
|
} |
282
|
|
|
$trace = 0.0; |
283
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
284
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
285
|
|
|
if ($i == $j) { |
286
|
|
|
$trace += $this->data[$i * $this->col + $i]; |
287
|
|
|
} |
288
|
|
|
} |
289
|
|
|
} |
290
|
|
|
return $trace; |
291
|
|
|
} |
292
|
|
|
|
293
|
|
|
/** |
294
|
|
|
* dignoalInterChange |
295
|
|
|
*/ |
296
|
|
|
public function dignoalInterChange() { |
297
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
298
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
299
|
|
|
$tmp = $this->data[$i * $this->col - $j]; |
300
|
|
|
$this->data[$i * $this->col - $j] = $tmp; |
301
|
|
|
} |
302
|
|
|
} |
303
|
|
|
} |
304
|
|
|
|
305
|
|
|
//---------------Arthmetic Opreations----------------------------------- |
306
|
|
|
|
307
|
|
|
/** |
308
|
|
|
* multiply this matrix with another matrix|scalar element-wise |
309
|
|
|
* Matrix Scalar\Matrix multiplication |
310
|
|
|
* @param int|float|matrix|vector $m |
311
|
|
|
* @return matrix|vector |
312
|
|
|
*/ |
313
|
|
|
public function multiply(int|float|matrix|vector $m): matrix|vector { |
314
|
|
|
if ($m instanceof self) { |
315
|
|
|
return $this->multiplyMatrix($m); |
316
|
|
|
} else if ($m instanceof vector) { |
317
|
|
|
return $this->multiplyVector($m); |
318
|
|
|
} else { |
319
|
|
|
return $this->scale($m); |
320
|
|
|
} |
321
|
|
|
} |
322
|
|
|
|
323
|
|
|
/** |
324
|
|
|
* |
325
|
|
|
* @param \Np\vector $v |
326
|
|
|
* @return matrix |
327
|
|
|
*/ |
328
|
|
|
protected function multiplyVector(vector $v): matrix { |
329
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this, $v)) { |
330
|
|
|
$ar = matrix::factory($this->row, $this->col, $this->dtype); |
331
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
332
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
333
|
|
|
$ar->data[$i * $this->col + $j] = $v->data[$j] * $this->data[$i * $this->col + $j]; |
334
|
|
|
} |
335
|
|
|
} |
336
|
|
|
return $ar; |
337
|
|
|
} |
338
|
|
|
} |
339
|
|
|
|
340
|
|
|
/** |
341
|
|
|
* |
342
|
|
|
* @param \Np\matrix $m |
343
|
|
|
* @return matrix |
344
|
|
|
*/ |
345
|
|
|
protected function multiplyMatrix(matrix $m): matrix { |
346
|
|
|
if ($this->checkDtype($this, $m) && $this->checkShape($this, $m)) { |
347
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
348
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
349
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
350
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] * $m->data[$i * $this->col + $j]; |
351
|
|
|
} |
352
|
|
|
} |
353
|
|
|
return $ar; |
354
|
|
|
} |
355
|
|
|
} |
356
|
|
|
|
357
|
|
|
/** |
358
|
|
|
* Sum of Rows of matrix |
359
|
|
|
* @return vector |
360
|
|
|
*/ |
361
|
|
|
public function sumRows(): vector { |
362
|
|
|
$vr = vector::factory($this->row, $this->dtype); |
363
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
364
|
|
|
$sum = 0.0; |
365
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
366
|
|
|
$sum += $this->data[$i * $this->col + $j]; |
367
|
|
|
} |
368
|
|
|
$vr->data[$i] = $sum; |
369
|
|
|
} |
370
|
|
|
return $vr; |
371
|
|
|
} |
372
|
|
|
|
373
|
|
|
/** |
374
|
|
|
* Sum of two matrix, vector or a scalar to current matrix |
375
|
|
|
* |
376
|
|
|
* @param int|float|matrix|vector $m |
377
|
|
|
* @return matrix |
378
|
|
|
*/ |
379
|
|
|
public function sum(int|float|matrix|vector $m): matrix { |
380
|
|
|
if ($m instanceof self) { |
381
|
|
|
return $this->sumMatrix($m); |
382
|
|
|
} elseif ($m instanceof vector) { |
383
|
|
|
return $this->sumVector($m); |
384
|
|
|
} else { |
385
|
|
|
return $this->sumScalar($m); |
386
|
|
|
} |
387
|
|
|
} |
388
|
|
|
|
389
|
|
|
protected function sumScalar(int|float $s): matrix { |
390
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
391
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
392
|
|
|
$ar->data[$i] = $this->data[$i] + $s; |
393
|
|
|
} |
394
|
|
|
return $ar; |
395
|
|
|
} |
396
|
|
|
|
397
|
|
|
protected function sumMatrix(matrix $m): matrix { |
398
|
|
|
if ($this->checkShape($this, $m) && $this->checkDtype($this, $m)) { |
399
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
400
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
401
|
|
|
$ar->data[$i] = $this->data[$i] + $m->data[$i]; |
402
|
|
|
} |
403
|
|
|
return $ar; |
404
|
|
|
} |
405
|
|
|
} |
406
|
|
|
|
407
|
|
|
protected function sumVector(vector $v): matrix { |
408
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this, $v)) { |
409
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
410
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
411
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
412
|
|
|
$ar->data[$i * $this->col + $j] = $v->data[$j] + $this->data[$i * $this->col + $j]; |
413
|
|
|
} |
414
|
|
|
} |
415
|
|
|
return $ar; |
416
|
|
|
} |
417
|
|
|
} |
418
|
|
|
|
419
|
|
|
/** |
420
|
|
|
* subtract another matrix, vector or a scalar to this matrix |
421
|
|
|
* @param int|float|matrix|vector $d matrix|$scalar to subtract this matrix |
422
|
|
|
* @return \Np\matrix |
423
|
|
|
*/ |
424
|
|
|
public function subtract(int|float|matrix|vector $d): matrix { |
425
|
|
|
if ($d instanceof self) { |
426
|
|
|
return $this->subtractMatrix($d); |
427
|
|
|
} elseif ($d instanceof vector) { |
428
|
|
|
return $this->subtractVector($d); |
429
|
|
|
} else { |
430
|
|
|
return $this->subtractScalar($d); |
431
|
|
|
} |
432
|
|
|
} |
433
|
|
|
|
434
|
|
|
protected function subtractScalar(int|float $s): matrix { |
435
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
436
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
437
|
|
|
$ar->data[$i] = $this->data[$i] - $s; |
438
|
|
|
} |
439
|
|
|
return $ar; |
440
|
|
|
} |
441
|
|
|
|
442
|
|
|
/** |
443
|
|
|
* |
444
|
|
|
* @param matrix $m |
445
|
|
|
* @return matrix |
446
|
|
|
*/ |
447
|
|
|
protected function subtractMatrix(matrix $m): matrix { |
448
|
|
|
if ($this->checkShape($this, $m) && $this->checkDtype($this, $m)) { |
449
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
450
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
451
|
|
|
$ar->data[$i] = $this->data[$i] - $m->data[$i]; |
452
|
|
|
} |
453
|
|
|
return $ar; |
454
|
|
|
} |
455
|
|
|
} |
456
|
|
|
|
457
|
|
|
/** |
458
|
|
|
* |
459
|
|
|
* @param vector $v |
460
|
|
|
* @return matrix |
461
|
|
|
*/ |
462
|
|
|
protected function subtractVector(vector $v): matrix { |
463
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this,$v)) { |
464
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
465
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
466
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
467
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] - $v->data[$j]; |
468
|
|
|
} |
469
|
|
|
} |
470
|
|
|
return $ar; |
471
|
|
|
} |
472
|
|
|
} |
473
|
|
|
|
474
|
|
|
/** |
475
|
|
|
* |
476
|
|
|
* @param vector $v |
477
|
|
|
* @return matrix |
478
|
|
|
*/ |
479
|
|
|
public function subtractColumnVector(vector $v): matrix { |
480
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this, $v)) { |
481
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
482
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
483
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
484
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] - $v->data[$i]; |
485
|
|
|
} |
486
|
|
|
} |
487
|
|
|
return $ar; |
488
|
|
|
} |
489
|
|
|
} |
490
|
|
|
|
491
|
|
|
/** |
492
|
|
|
* Return the division of two elements, element-wise. |
493
|
|
|
* @param int|float|matrix $d |
494
|
|
|
* @return matrix |
495
|
|
|
*/ |
496
|
|
|
public function divide(int|float|matrix|vector $d): matrix { |
497
|
|
|
if ($d instanceof self) { |
498
|
|
|
return $this->divideMatrix($d); |
499
|
|
|
} elseif ($d instanceof vector) { |
500
|
|
|
return $this->divideVector($d); |
501
|
|
|
} else { |
502
|
|
|
return $this->divideScalar($d); |
503
|
|
|
} |
504
|
|
|
} |
505
|
|
|
|
506
|
|
|
protected function divideMatrix(matrix $m): matrix { |
507
|
|
|
if ($this->checkShape($this, $m) && $this->checkDtype($this, $m)) { |
508
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
509
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
510
|
|
|
$ar->data[$i] = $this->data[$i] / $m->data[$i]; |
511
|
|
|
} |
512
|
|
|
return $ar; |
513
|
|
|
} |
514
|
|
|
} |
515
|
|
|
|
516
|
|
|
protected function divideVector(vector $v): matrix { |
517
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this, $v)) { |
518
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
519
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
520
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
521
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] / $v->data[$j]; |
522
|
|
|
} |
523
|
|
|
} |
524
|
|
|
return $ar; |
525
|
|
|
} |
526
|
|
|
} |
527
|
|
|
|
528
|
|
|
protected function divideScalar(int|float $s): matrix { |
529
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
530
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
531
|
|
|
$ar->data[$i] = $this->data[$i] / $s; |
532
|
|
|
} |
533
|
|
|
return $ar; |
534
|
|
|
} |
535
|
|
|
|
536
|
|
|
/** |
537
|
|
|
* |
538
|
|
|
* Raise this matrix to the power of the element-wise entry in another matrix. |
539
|
|
|
* |
540
|
|
|
* @param int|float|matrix $m |
541
|
|
|
* @return matrix |
542
|
|
|
*/ |
543
|
|
|
public function pow(int|float|matrix|vector $d): matrix { |
544
|
|
|
if ($d instanceof self) { |
545
|
|
|
return $this->powMatrix($d); |
546
|
|
|
} else if ($d instanceof vector) { |
547
|
|
|
return $this->powVector($d); |
548
|
|
|
} else { |
549
|
|
|
return $this->powScalar($d); |
550
|
|
|
} |
551
|
|
|
} |
552
|
|
|
|
553
|
|
|
protected function powMatrix(matrix $m): matrix { |
554
|
|
|
if ($this->checkShape($this, $m) && $this->checkDtype($this, $m)) { |
555
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
556
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
557
|
|
|
$ar->data[$i] = $this->data[$i] ** $m->data[$i]; |
558
|
|
|
} |
559
|
|
|
return $ar; |
560
|
|
|
} |
561
|
|
|
} |
562
|
|
|
|
563
|
|
|
protected function powVector(vector $v): matrix { |
564
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this, $v)) { |
565
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
566
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
567
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
568
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] ** $v->data[$j]; |
569
|
|
|
} |
570
|
|
|
} |
571
|
|
|
return $ar; |
572
|
|
|
} |
573
|
|
|
} |
574
|
|
|
|
575
|
|
|
protected function powScalar(int|float $s): matrix { |
576
|
|
|
$ar = $this->copy(); |
577
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
578
|
|
|
$ar->data[$i] **= $s; |
579
|
|
|
} |
580
|
|
|
return $ar; |
581
|
|
|
} |
582
|
|
|
|
583
|
|
|
/** |
584
|
|
|
* Calculate the modulus i.e remainder of division between this matrix and another matrix. |
585
|
|
|
* @param int|float|matrix|vector $d |
586
|
|
|
* @return matrix |
587
|
|
|
*/ |
588
|
|
|
public function mod(int|float|matrix|vector $d): matrix { |
589
|
|
|
if ($d instanceof self) { |
590
|
|
|
$this->modMatrix($d); |
591
|
|
|
} else if ($d instanceof vector) { |
592
|
|
|
$this->modVector($d); |
593
|
|
|
} else { |
594
|
|
|
$this->modScalar($d); |
595
|
|
|
} |
596
|
|
|
} |
597
|
|
|
|
598
|
|
|
protected function modMatrix(matrix $m): matrix { |
599
|
|
|
if ($this->checkShape($this, $m) && $this->checkDtype($this, $m)) { |
600
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
601
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
602
|
|
|
$ar->data[$i] = $this->data[$i] % $m->data[$i]; |
603
|
|
|
} |
604
|
|
|
return $ar; |
605
|
|
|
} |
606
|
|
|
} |
607
|
|
|
|
608
|
|
|
protected function modVector(vector $v): matrix { |
609
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this, $v)) { |
610
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
611
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
612
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
613
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] % $v->data[$j]; |
614
|
|
|
} |
615
|
|
|
} |
616
|
|
|
return $ar; |
617
|
|
|
} |
618
|
|
|
} |
619
|
|
|
|
620
|
|
|
protected function modScalar(int|float $s): matrix { |
621
|
|
|
$ar = $this->copy(); |
622
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
623
|
|
|
$ar->data[$i] %= $s; |
624
|
|
|
} |
625
|
|
|
return $ar; |
626
|
|
|
} |
627
|
|
|
|
628
|
|
|
/** |
629
|
|
|
* Return the element-wise reciprocal of the matrix. |
630
|
|
|
* |
631
|
|
|
* @return matrix |
632
|
|
|
*/ |
633
|
|
|
public function reciprocal(): matrix { |
634
|
|
|
return self::ones($this->row, $this->col, $this->dtype)->divideMatrix($this); |
635
|
|
|
} |
636
|
|
|
|
637
|
|
|
/** |
638
|
|
|
* |
639
|
|
|
* @param int|float $d |
640
|
|
|
* @return bool |
641
|
|
|
*/ |
642
|
|
|
public static function is_zero($d): bool { |
643
|
|
|
if (abs($d) < self::EPSILON) { |
644
|
|
|
return true; |
645
|
|
|
} |
646
|
|
|
return false; |
647
|
|
|
} |
648
|
|
|
|
649
|
|
|
/** |
650
|
|
|
* is row zero |
651
|
|
|
* @param int $row |
652
|
|
|
* @return bool |
653
|
|
|
*/ |
654
|
|
|
public function is_rowZero(int $row): bool { |
655
|
|
|
for ($i = 0; $i < $this->col; ++$i) { |
656
|
|
|
if ($this->data[$row * $this->col + $i] != 0) { |
657
|
|
|
return false; |
658
|
|
|
} |
659
|
|
|
} |
660
|
|
|
return true; |
661
|
|
|
} |
662
|
|
|
|
663
|
|
|
/** |
664
|
|
|
* |
665
|
|
|
* @return bool |
666
|
|
|
*/ |
667
|
|
|
public function has_ZeroRow(): bool { |
668
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
669
|
|
|
if ($this->is_rowZero($i)) { |
670
|
|
|
return true; |
671
|
|
|
} |
672
|
|
|
} |
673
|
|
|
return false; |
674
|
|
|
} |
675
|
|
|
|
676
|
|
|
/** |
677
|
|
|
* Transpose the matrix i.e row become cols and cols become rows. |
678
|
|
|
* @return \Np\matrix |
679
|
|
|
*/ |
680
|
|
|
public function transpose(): matrix { |
681
|
|
|
$ar = self::factory($this->col, $this->row, $this->dtype); |
682
|
|
|
for ($i = 0; $i < $ar->row; ++$i) { |
683
|
|
|
for ($j = 0; $j < $ar->col; ++$j) { |
684
|
|
|
$ar->data[$i * $ar->col + $j] = $this->data[$j * $ar->col + $i]; |
685
|
|
|
} |
686
|
|
|
} |
687
|
|
|
return $ar; |
688
|
|
|
} |
689
|
|
|
|
690
|
|
|
/** |
691
|
|
|
* swap specific values in matrix |
692
|
|
|
* @param int $i1 |
693
|
|
|
* @param int $i2 |
694
|
|
|
*/ |
695
|
|
|
public function swapValue(int $i1, int $i2) { |
696
|
|
|
$tmp = $this->data[$i1]; |
697
|
|
|
$this->data[$i1] = $this->data[$i2]; |
698
|
|
|
$this->data[$i2] = $tmp; |
699
|
|
|
} |
700
|
|
|
|
701
|
|
|
/** |
702
|
|
|
* swap specific rows in matrix |
703
|
|
|
* @param int $r1 |
704
|
|
|
* @param int $r2 |
705
|
|
|
*/ |
706
|
|
|
public function swapRows(int $r1, int $r2) { |
707
|
|
|
for ($i = 0; $i < $this->col; ++$i) { |
708
|
|
|
$tmp = $this->data[$r1 * $this->col + $i]; |
709
|
|
|
$this->data[$r1 * $this->col + $i] = $this->data[$r2 * $this->col + $i]; |
710
|
|
|
$this->data[$r2 * $this->col + $i] = $tmp; |
711
|
|
|
} |
712
|
|
|
} |
713
|
|
|
|
714
|
|
|
/** |
715
|
|
|
* swap specific cols in matrix |
716
|
|
|
* @param int $c1 |
717
|
|
|
* @param int $c2 |
718
|
|
|
*/ |
719
|
|
|
public function swapCols(int $c1, int $c2) { |
720
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
721
|
|
|
$tmp = $this->data[$i * $this->row + $c1]; |
722
|
|
|
$this->data[$i * $this->row + $c1] = $this->data[$i * $this->row + $c2]; |
723
|
|
|
$this->data[$i * $this->row + $c2] = $tmp; |
724
|
|
|
} |
725
|
|
|
} |
726
|
|
|
|
727
|
|
|
/** |
728
|
|
|
* |
729
|
|
|
* @param int|float $scalar |
730
|
|
|
* @return matrix |
731
|
|
|
*/ |
732
|
|
|
public function scale(int|float $scalar): matrix { |
733
|
|
|
if ($scalar == 0) { |
734
|
|
|
return self::zeros($this->row, $this->col, $this->dtype); |
735
|
|
|
} |
736
|
|
|
|
737
|
|
|
$ar = $this->copy(); |
738
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
739
|
|
|
$ar->data[$i] *= $scalar; |
740
|
|
|
} |
741
|
|
|
|
742
|
|
|
return $ar; |
743
|
|
|
} |
744
|
|
|
|
745
|
|
|
/** |
746
|
|
|
* scale all the elements of a row |
747
|
|
|
* @param int $row |
748
|
|
|
* @param int|float $c |
749
|
|
|
*/ |
750
|
|
|
public function scaleRow(int $row, int|float $c) { |
751
|
|
|
for ($i = 0; $i < $this->col; ++$i) { |
752
|
|
|
$this->data[$row * $this->col + $i] *= $c; |
753
|
|
|
} |
754
|
|
|
} |
755
|
|
|
|
756
|
|
|
/** |
757
|
|
|
* scale all the elements of |
758
|
|
|
* @param int $col |
759
|
|
|
* @param int|float $c |
760
|
|
|
*/ |
761
|
|
|
public function scaleCol(int $col, int|float $c) { |
762
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
763
|
|
|
$this->data[$i * $this->col + $col] *= $c; |
764
|
|
|
} |
765
|
|
|
} |
766
|
|
|
|
767
|
|
|
/** |
768
|
|
|
* Scale digonally |
769
|
|
|
* @param int|float $c |
770
|
|
|
* @param bool $lDig |
771
|
|
|
*/ |
772
|
|
|
public function scaleDigonalCol(int|float $c, bool $lDig = true) { |
773
|
|
|
if($lDig){ |
774
|
|
|
for ($i = 0; $i < $this->row ; ++$i) { |
775
|
|
|
$this->data[$i * $this->col + $i] *= $c; |
776
|
|
|
} |
777
|
|
|
} |
778
|
|
|
else{ |
779
|
|
|
for ($i = $this->row; $i > 0 ; --$i) { |
780
|
|
|
$this->data[$i * $this->col - $i] *= $c; |
781
|
|
|
} |
782
|
|
|
} |
783
|
|
|
} |
784
|
|
|
|
785
|
|
|
/** |
786
|
|
|
* |
787
|
|
|
* @param int $r1 |
788
|
|
|
* @param int $r2 |
789
|
|
|
* @param float $c |
790
|
|
|
*/ |
791
|
|
|
public function addScaleRow(int $r1, int $r2, float $c) { |
792
|
|
|
for ($i = 0; $i < $this->col; ++$i) { |
793
|
|
|
$this->data[$r2 * $this->col + $i] += $this->data[$r1 * $this->col + $i] * $c; |
794
|
|
|
} |
795
|
|
|
} |
796
|
|
|
|
797
|
|
|
/** |
798
|
|
|
* Attach given matrix to the left of this matrix. |
799
|
|
|
* |
800
|
|
|
* @param \Np\matrix $m |
801
|
|
|
* @return \Np\matrix |
802
|
|
|
*/ |
803
|
|
|
public function joinLeft(matrix $m): matrix { |
804
|
|
|
if ($this->row != $m->row && !$this->checkDtype($this, $m)) { |
805
|
|
|
self::_err('Error::Invalid size! or DataType!'); |
806
|
|
|
} |
807
|
|
|
$col = $this->col + $m->col; |
808
|
|
|
$ar = self::factory($this->row, $col, $this->dtype); |
809
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
810
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
811
|
|
|
$ar->data[$i * $col + $j] = $this->data[$i * $this->col + $j]; |
812
|
|
|
} |
813
|
|
|
for ($j = 0; $j < $m->col; ++$j) { |
814
|
|
|
$ar->data[$i * $col + ($this->col + $j)] = $m->data[$i * $m->col + $j]; |
815
|
|
|
} |
816
|
|
|
} |
817
|
|
|
return $ar; |
818
|
|
|
} |
819
|
|
|
|
820
|
|
|
/** |
821
|
|
|
* Join matrix m to the Right of this matrix. |
822
|
|
|
* @param \Np\matrix $m |
823
|
|
|
* @return matrix |
824
|
|
|
*/ |
825
|
|
|
public function joinRight(matrix $m): matrix { |
826
|
|
|
if ($this->row != $m->row && !$this->checkDtype($this,$m)) { |
827
|
|
|
self::_err('Error::Invalid size! or DataType!'); |
828
|
|
|
} |
829
|
|
|
$col = $this->col + $m->col; |
830
|
|
|
$ar = self::factory($this->row, $col, $this->dtype); |
831
|
|
|
for ($i = 0; $i < $m->row; ++$i) { |
832
|
|
|
for ($j = 0; $j < $m->col; ++$j) { |
833
|
|
|
$ar->data[$i * $col + $j] = $m->data[$i * $m->col + $j]; |
834
|
|
|
} |
835
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
836
|
|
|
$ar->data[$i * $col + ($this->col + $j)] = $this->data[$i * $this->col + $j]; |
837
|
|
|
} |
838
|
|
|
} |
839
|
|
|
return $ar; |
840
|
|
|
} |
841
|
|
|
|
842
|
|
|
/** |
843
|
|
|
* Join matrix m Above this matrix. |
844
|
|
|
* @param \Np\matrix $m |
845
|
|
|
* @return matrix |
846
|
|
|
*/ |
847
|
|
|
public function joinAbove(matrix $m): matrix { |
848
|
|
|
if ($this->col !== $m->col && !$this->checkDtype($this, $m)) { |
849
|
|
|
self::_err('Error::Invalid size! or DataType!'); |
850
|
|
|
} |
851
|
|
|
$row = $this->row + $m->row; |
852
|
|
|
$ar = self::factory($row, $this->col, $this->dtype); |
853
|
|
|
for ($i = 0; $i < $m->row; ++$i) { |
854
|
|
|
for ($j = 0; $j < $m->col; ++$j) { |
855
|
|
|
$ar->data[$i * $m->col + $j] = $m->data[$i * $m->col + $j]; |
856
|
|
|
} |
857
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
858
|
|
|
$ar->data[($i + $this->row) * $this->col + $j] = $this->data[$i * $this->col + $j]; |
859
|
|
|
} |
860
|
|
|
} |
861
|
|
|
return $ar; |
862
|
|
|
} |
863
|
|
|
|
864
|
|
|
/** |
865
|
|
|
* Join matrix m below this matrix. |
866
|
|
|
* @param \Np\matrix $m |
867
|
|
|
* @return matrix |
868
|
|
|
*/ |
869
|
|
|
public function joinBelow(matrix $m): matrix { |
870
|
|
|
if ($this->col !== $m->col && !$this->checkDtype($this, $m)) { |
871
|
|
|
self::_err('Error::Invalid size! or DataType!'); |
872
|
|
|
} |
873
|
|
|
$row = $this->row + $m->row; |
874
|
|
|
$ar = self::factory($row, $this->col, $this->dtype); |
875
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
876
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
877
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j]; |
878
|
|
|
} |
879
|
|
|
for ($j = 0; $j < $m->col; ++$j) { |
880
|
|
|
$ar->data[($i + $m->row) * $m->col + $j] = $m->data[$i * $m->col + $j]; |
881
|
|
|
} |
882
|
|
|
} |
883
|
|
|
return $ar; |
884
|
|
|
} |
885
|
|
|
|
886
|
|
|
/** |
887
|
|
|
* Calculate the row echelon form of the matrix. |
888
|
|
|
* Return the reduced matrix. |
889
|
|
|
* |
890
|
|
|
* @return matrix|null |
891
|
|
|
*/ |
892
|
|
|
public function ref(): matrix|null { |
893
|
|
|
return ref::factory($this); |
894
|
|
|
} |
895
|
|
|
|
896
|
|
|
/** |
897
|
|
|
* Return the lower triangular matrix of the Cholesky decomposition. |
898
|
|
|
* |
899
|
|
|
* @return matrix|null |
900
|
|
|
*/ |
901
|
|
|
public function cholesky(): matrix|null { |
902
|
|
|
return cholesky::factory($this); |
903
|
|
|
} |
904
|
|
|
|
905
|
|
|
/** |
906
|
|
|
* FIXME-------------- |
907
|
|
|
* RREF |
908
|
|
|
* The reduced row echelon form (RREF) of a matrix. |
909
|
|
|
* @return \Np\matrix |
910
|
|
|
*/ |
911
|
|
|
public function rref(): matrix { |
912
|
|
|
return rref::factory($this); |
913
|
|
|
} |
914
|
|
|
|
915
|
|
|
/** |
916
|
|
|
* |
917
|
|
|
* @param int $cols |
918
|
|
|
* @return \Np\matrix |
919
|
|
|
*/ |
920
|
|
|
public function diminish_left(int $cols): matrix { |
921
|
|
|
$ar = self::factory($this->row, $cols, $this->dtype); |
922
|
|
|
for ($i = 0; $i < $ar->row; ++$i) { |
923
|
|
|
for ($j = 0; $j < $ar->col; ++$j) { |
924
|
|
|
$ar->data[$i * $ar->col + $j] = $this->data[$i * $this->col + $j]; |
925
|
|
|
} |
926
|
|
|
} |
927
|
|
|
return $ar; |
928
|
|
|
} |
929
|
|
|
|
930
|
|
|
/** |
931
|
|
|
* |
932
|
|
|
* @param int $cols |
933
|
|
|
* @return \Np\matrix |
934
|
|
|
*/ |
935
|
|
|
public function diminish_right(int $cols): matrix { |
936
|
|
|
$ar = self::factory($this->row, $cols, $this->dtype); |
937
|
|
|
for ($i = 0; $i < $ar->row; ++$i) { |
938
|
|
|
for ($j = 0; $j < $ar->col; ++$j) { |
939
|
|
|
$ar->data[$i * $ar->col + $j] = $this->data[$i * $this->col - $cols + $j]; |
940
|
|
|
} |
941
|
|
|
} |
942
|
|
|
return $ar; |
943
|
|
|
} |
944
|
|
|
|
945
|
|
|
/** |
946
|
|
|
* Return the index of the maximum element in every row of the matrix. |
947
|
|
|
* @return \Np\vector int |
948
|
|
|
*/ |
949
|
|
|
public function argMax(): vector { |
950
|
|
|
$v = vector::factory($this->row, vector::INT); |
951
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
952
|
|
|
$v->data[$i] = blas::max($this->rowAsVector($i)); |
953
|
|
|
} |
954
|
|
|
return $v; |
955
|
|
|
} |
956
|
|
|
|
957
|
|
|
/** |
958
|
|
|
* Return the index of the minimum element in every row of the matrix. |
959
|
|
|
* @return \Np\vector int |
960
|
|
|
*/ |
961
|
|
|
public function argMin(): vector { |
962
|
|
|
$v = vector::factory($this->row, vector::INT); |
963
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
964
|
|
|
$v->data[$i] = blas::min($this->rowAsVector($i)); |
965
|
|
|
} |
966
|
|
|
|
967
|
|
|
return $v; |
968
|
|
|
} |
969
|
|
|
|
970
|
|
|
/** |
971
|
|
|
* Set given data in matrix |
972
|
|
|
* @param int|float|array $data |
973
|
|
|
* @param bool $dignoal |
974
|
|
|
* @return void |
975
|
|
|
*/ |
976
|
|
|
public function setData(int|float|array $data, bool $dignoal = false): void { |
977
|
|
|
if ($dignoal == false) { |
978
|
|
|
if (is_array($data) && is_array($data[0])) { |
979
|
|
|
$f = $this->flattenArray($data); |
980
|
|
|
foreach ($f as $k => $v) { |
981
|
|
|
$this->data[$k] = $v; |
982
|
|
|
} |
983
|
|
|
} elseif (is_numeric($data)) { |
984
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
985
|
|
|
$this->data[$i] = $data; |
986
|
|
|
} |
987
|
|
|
} |
988
|
|
|
} elseif (is_numeric($data) || is_array($data) && !is_array($data[0])) { |
989
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
990
|
|
|
$this->data[$i * $this->col * $i] = $data; |
991
|
|
|
} |
992
|
|
|
} |
993
|
|
|
} |
994
|
|
|
|
995
|
|
|
/** |
996
|
|
|
* get the matrix data type |
997
|
|
|
* @return int |
998
|
|
|
*/ |
999
|
|
|
public function getDtype():int { |
1000
|
|
|
return $this->dtype; |
1001
|
|
|
} |
1002
|
|
|
|
1003
|
|
|
/** |
1004
|
|
|
* get the shape of matrix |
1005
|
|
|
* @return object |
1006
|
|
|
*/ |
1007
|
|
|
public function getShape(): object { |
1008
|
|
|
return (object) ['m' => $this->row, 'n' => $this->col]; |
1009
|
|
|
} |
1010
|
|
|
|
1011
|
|
|
/** |
1012
|
|
|
* get the number of elements in the matrix. |
1013
|
|
|
* @return int |
1014
|
|
|
*/ |
1015
|
|
|
public function getSize(): int { |
1016
|
|
|
return $this->ndim; |
1017
|
|
|
} |
1018
|
|
|
|
1019
|
|
|
/** |
1020
|
|
|
* is matrix squred |
1021
|
|
|
* @return bool |
1022
|
|
|
*/ |
1023
|
|
|
public function isSquare(): bool { |
1024
|
|
|
if ($this->row === $this->col) { |
1025
|
|
|
return true; |
1026
|
|
|
} |
1027
|
|
|
return false; |
1028
|
|
|
} |
1029
|
|
|
|
1030
|
|
|
/** |
1031
|
|
|
* Return a row as vector from the matrix. |
1032
|
|
|
* @param int $index |
1033
|
|
|
* @return \Np\vector |
1034
|
|
|
*/ |
1035
|
|
|
public function rowAsVector(int $index): vector { |
1036
|
|
|
$vr = vector::factory($this->col, $this->dtype); |
1037
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
1038
|
|
|
$vr->data[$j] = $this->data[$index * $this->col + $j]; |
1039
|
|
|
} |
1040
|
|
|
return $vr; |
1041
|
|
|
} |
1042
|
|
|
|
1043
|
|
|
/** |
1044
|
|
|
* Return a col as vector from the matrix. |
1045
|
|
|
* @param int $index |
1046
|
|
|
* @return \Np\vector |
1047
|
|
|
*/ |
1048
|
|
|
public function colAsVector(int $index): vector { |
1049
|
|
|
$vr = vector::factory($this->row, $this->dtype); |
1050
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
1051
|
|
|
$vr->data[$i] = $this->data[$i * $this->row + $index]; |
1052
|
|
|
} |
1053
|
|
|
return $vr; |
1054
|
|
|
} |
1055
|
|
|
|
1056
|
|
|
/** |
1057
|
|
|
* Return the diagonal elements of a square matrix as a vector. |
1058
|
|
|
* @return \Np\vector |
1059
|
|
|
*/ |
1060
|
|
|
public function diagonalAsVector(): vector { |
1061
|
|
|
if (!$this->isSquare()) { |
1062
|
|
|
self::_err('Can not trace of a none square matrix'); |
1063
|
|
|
} |
1064
|
|
|
$vr = vector::factory($this->row, $this->dtype); |
1065
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
1066
|
|
|
$vr->data[$i] = $this->getDiagonalVal($i); |
1067
|
|
|
} |
1068
|
|
|
return $vr; |
1069
|
|
|
} |
1070
|
|
|
|
1071
|
|
|
/** |
1072
|
|
|
* Flatten i.e unravel the matrix into a vector. |
1073
|
|
|
* |
1074
|
|
|
* @return \Np\vector |
1075
|
|
|
*/ |
1076
|
|
|
public function asVector(): vector { |
1077
|
|
|
$vr = vector::factory($this->ndim, $this->dtype); |
1078
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
1079
|
|
|
$vr->data[$i] = $this->data[$i]; |
1080
|
|
|
} |
1081
|
|
|
return $vr; |
1082
|
|
|
} |
1083
|
|
|
|
1084
|
|
|
/** |
1085
|
|
|
* Return the elements of the matrix in a 2-d array. |
1086
|
|
|
* @return array |
1087
|
|
|
*/ |
1088
|
|
|
public function asArray(): array { |
1089
|
|
|
$ar = array_fill(0, $this->row, array_fill(0, $this->col, null)); |
1090
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
1091
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
1092
|
|
|
$ar[$i][$j] = $this->data[$i * $this->col + $j]; |
1093
|
|
|
} |
1094
|
|
|
} |
1095
|
|
|
return $ar; |
1096
|
|
|
} |
1097
|
|
|
|
1098
|
|
|
/** |
1099
|
|
|
* get a diagonal value from matrix |
1100
|
|
|
* @param int $i |
1101
|
|
|
* @return float |
1102
|
|
|
*/ |
1103
|
|
|
public function getDiagonalVal(int $i) { |
1104
|
|
|
if ($this->isSquare()) { |
1105
|
|
|
return $this->data[$i * $this->row + $i]; |
1106
|
|
|
} |
1107
|
|
|
} |
1108
|
|
|
|
1109
|
|
|
|
1110
|
|
|
|
1111
|
|
|
/** |
1112
|
|
|
* Compute the singular value decomposition of a matrix and |
1113
|
|
|
* return an object of the singular values and unitary matrices |
1114
|
|
|
* |
1115
|
|
|
* @return object (u,s,v) |
1116
|
|
|
*/ |
1117
|
|
|
public function svd(): svd { |
1118
|
|
|
return svd::factory($this); |
1119
|
|
|
} |
1120
|
|
|
|
1121
|
|
|
/** |
1122
|
|
|
* Compute the eigen decomposition of a general matrix. |
1123
|
|
|
* return the eigenvalues and eigenvectors as object |
1124
|
|
|
* |
1125
|
|
|
* @param bool $symmetric |
1126
|
|
|
* @return eigen |
1127
|
|
|
*/ |
1128
|
|
|
public function eign(bool $symmetric = false): eigen { |
1129
|
|
|
return eigen::factory($this, $symmetric); |
1130
|
|
|
} |
1131
|
|
|
|
1132
|
|
|
/** |
1133
|
|
|
* |
1134
|
|
|
* Compute the LU factorization of matrix. |
1135
|
|
|
* return lower, upper, and permutation matrices as object. |
1136
|
|
|
* |
1137
|
|
|
* @return lu |
1138
|
|
|
*/ |
1139
|
|
|
public function lu(): lu { |
1140
|
|
|
return lu::factory($this); |
1141
|
|
|
} |
1142
|
|
|
|
1143
|
|
|
/** |
1144
|
|
|
* Return the L1 norm of the matrix. |
1145
|
|
|
* @return float |
1146
|
|
|
*/ |
1147
|
|
|
public function normL1(): float { |
1148
|
|
|
return lapack::lange('l', $this); |
1149
|
|
|
} |
1150
|
|
|
|
1151
|
|
|
/** |
1152
|
|
|
* Return the L2 norm of the matrix. |
1153
|
|
|
* @return float |
1154
|
|
|
*/ |
1155
|
|
|
public function normL2(): float { |
1156
|
|
|
return lapack::lange('f', $this); |
1157
|
|
|
} |
1158
|
|
|
|
1159
|
|
|
/** |
1160
|
|
|
* Return the L1 norm of the matrix. |
1161
|
|
|
* @return float |
1162
|
|
|
*/ |
1163
|
|
|
public function normINF(): float { |
1164
|
|
|
return lapack::lange('i', $this); |
1165
|
|
|
} |
1166
|
|
|
|
1167
|
|
|
/** |
1168
|
|
|
* Return the Frobenius norm of the matrix. |
1169
|
|
|
* @return float |
1170
|
|
|
*/ |
1171
|
|
|
public function normFrob(): float { |
1172
|
|
|
return $this->normL2(); |
1173
|
|
|
} |
1174
|
|
|
|
1175
|
|
|
/** |
1176
|
|
|
* Compute the means of each row and return them in a vector. |
1177
|
|
|
* |
1178
|
|
|
* @return vector |
1179
|
|
|
*/ |
1180
|
|
|
public function mean(): vector { |
1181
|
|
|
return $this->sumRows()->divide($this->col); |
1182
|
|
|
} |
1183
|
|
|
|
1184
|
|
|
/** |
1185
|
|
|
* Compute the row variance of the matrix. |
1186
|
|
|
* |
1187
|
|
|
* @param vector|null $mean |
1188
|
|
|
* @return vector |
1189
|
|
|
*/ |
1190
|
|
|
public function variance(vector|null $mean = null): vector { |
1191
|
|
|
if (isset($mean)) { |
1192
|
|
|
if (!$mean instanceof vector) { |
1193
|
|
|
self::_invalidArgument('mean must be a vector!'); |
1194
|
|
|
} |
1195
|
|
|
if ($this->row !== $mean->col) { |
1196
|
|
|
self::_err('Err:: given mean vector dimensionality mismatched!'); |
1197
|
|
|
} |
1198
|
|
|
} else { |
1199
|
|
|
$mean = $this->mean(); |
1200
|
|
|
} |
1201
|
|
|
return $this->subtractColumnVector($mean)->square() |
1202
|
|
|
->sumRows()->divide($this->row); |
1203
|
|
|
} |
1204
|
|
|
|
1205
|
|
|
/** |
1206
|
|
|
* Return the median vector of this matrix. |
1207
|
|
|
* @return vector |
1208
|
|
|
*/ |
1209
|
|
|
public function median(): vector { |
1210
|
|
|
$mid = intdiv($this->col, 2); |
1211
|
|
|
$odd = $this->col % 2 === 1; |
1212
|
|
|
$vr = vector::factory($this->row, $this->dtype); |
1213
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
1214
|
|
|
$a = $this->rowAsVector($i)->sort(); |
1215
|
|
|
if ($odd) { |
1216
|
|
|
$median = $a->data[$mid]; |
1217
|
|
|
} else { |
1218
|
|
|
$median = ($a->data[$mid - 1] + $a->data[$mid]) / 2.0; |
1219
|
|
|
} |
1220
|
|
|
$vr->data[$i] = $median; |
1221
|
|
|
} |
1222
|
|
|
unset($a); |
1223
|
|
|
return $vr; |
1224
|
|
|
} |
1225
|
|
|
|
1226
|
|
|
/** |
1227
|
|
|
* Compute the covariance matrix. |
1228
|
|
|
* |
1229
|
|
|
* @param vector|null $mean |
1230
|
|
|
* @return matrix |
1231
|
|
|
*/ |
1232
|
|
|
public function covariance(vector|null $mean = null): matrix { |
1233
|
|
|
if (isset($mean)) { |
1234
|
|
|
if ($mean->col !== $this->row) { |
1235
|
|
|
self::_err('Err:: given mean vector dimensionality mismatched!'); |
1236
|
|
|
} |
1237
|
|
|
} else { |
1238
|
|
|
$mean = $this->mean(); |
1239
|
|
|
} |
1240
|
|
|
|
1241
|
|
|
$b = $this->subtractColumnVector($mean); |
1242
|
|
|
|
1243
|
|
|
return $b->dot($b->transpose()) |
1244
|
|
|
->divideScalar($this->row); |
1245
|
|
|
} |
1246
|
|
|
|
1247
|
|
|
/** |
1248
|
|
|
* Square of matrix |
1249
|
|
|
* @return matrix |
1250
|
|
|
*/ |
1251
|
|
|
public function square(): matrix { |
1252
|
|
|
return $this->multiplyMatrix($this); |
1253
|
|
|
} |
1254
|
|
|
|
1255
|
|
|
/** |
1256
|
|
|
* |
1257
|
|
|
* @param int|float|matrix|vector $d |
1258
|
|
|
* @return matrix |
1259
|
|
|
*/ |
1260
|
|
|
public function equal(int|float|matrix|vector $d): matrix { |
1261
|
|
|
if ($d instanceof self) { |
1262
|
|
|
return $this->equalMatrix($d); |
1263
|
|
|
} elseif ($d instanceof vector) { |
1264
|
|
|
return $this->equalVector($d); |
1265
|
|
|
} else { |
1266
|
|
|
return $this->equalScalar($d); |
1267
|
|
|
} |
1268
|
|
|
} |
1269
|
|
|
|
1270
|
|
|
protected function equalMatrix(matrix $m): matrix { |
1271
|
|
|
if ($this->checkShape($this, $m) && $this->checkDtype($this, $m)) { |
1272
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
1273
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
1274
|
|
|
$ar->data[$i] = $this->data[$i] == $m->data[$i] ? 1 : 0; |
1275
|
|
|
} |
1276
|
|
|
return $ar; |
1277
|
|
|
} |
1278
|
|
|
} |
1279
|
|
|
|
1280
|
|
|
protected function equalVector(vector $v): matrix { |
1281
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this, $v)) { |
1282
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
1283
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
1284
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
1285
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] == $v->data[$j] ? 1 : 0; |
1286
|
|
|
} |
1287
|
|
|
} |
1288
|
|
|
return $ar; |
1289
|
|
|
} |
1290
|
|
|
} |
1291
|
|
|
|
1292
|
|
|
protected function equalScalar(int|float $s): matrix { |
1293
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
1294
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
1295
|
|
|
$ar->data[$i] = $this->data[$i] == $s ? 1 : 0; |
1296
|
|
|
} |
1297
|
|
|
return $ar; |
1298
|
|
|
} |
1299
|
|
|
|
1300
|
|
|
/** |
1301
|
|
|
* |
1302
|
|
|
* @param int|float|matrix|vector $d |
1303
|
|
|
* @return matrix |
1304
|
|
|
*/ |
1305
|
|
|
public function greater(int|float|matrix|vector $d): matrix { |
1306
|
|
|
if ($d instanceof self) { |
1307
|
|
|
return $this->greaterMatrix($d); |
1308
|
|
|
} elseif ($d instanceof vector) { |
1309
|
|
|
return $this->greaterVector($d); |
1310
|
|
|
} else { |
1311
|
|
|
return $this->greaterScalar($d); |
1312
|
|
|
} |
1313
|
|
|
} |
1314
|
|
|
|
1315
|
|
|
protected function greaterMatrix(matrix $m): matrix { |
1316
|
|
|
if ($this->checkShape($this, $m) && $this->checkDtype($this,$m)) { |
1317
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
1318
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
1319
|
|
|
$ar->data[$i] = $this->data[$i] > $m->data[$i] ? 1 : 0; |
1320
|
|
|
} |
1321
|
|
|
return $ar; |
1322
|
|
|
} |
1323
|
|
|
} |
1324
|
|
|
|
1325
|
|
|
protected function greaterVector(vector $v): matrix { |
1326
|
|
|
if ($this->checkDimensions($v, $this) && $this->checkDtype($this,$v)) { |
1327
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
1328
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
1329
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
1330
|
|
|
$ar->data[$i * $this->col + $j] = $this->data[$i * $this->col + $j] > $v->data[$j] ? 1 : 0; |
1331
|
|
|
} |
1332
|
|
|
} |
1333
|
|
|
return $ar; |
1334
|
|
|
} |
1335
|
|
|
} |
1336
|
|
|
|
1337
|
|
|
protected function greaterScalar(int|float $s): matrix { |
1338
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
1339
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
1340
|
|
|
$ar->data[$i] = $this->data[$i] > $s ? 1 : 0; |
1341
|
|
|
} |
1342
|
|
|
return $ar; |
1343
|
|
|
} |
1344
|
|
|
|
1345
|
|
|
/** |
1346
|
|
|
* |
1347
|
|
|
* @param int|float|matrix $m |
1348
|
|
|
* @return matrix |
1349
|
|
|
*/ |
1350
|
|
|
public function less(int|float|matrix $m): matrix { |
1351
|
|
|
$ar = self::factory($this->row, $this->col, $this->dtype); |
1352
|
|
|
if ($m instanceof self) { |
1353
|
|
|
if ($this->checkShape($this,$m)) { |
1354
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
1355
|
|
|
$ar->data[$i] = $this->data[$i] < $m->data[$i] ? 1 : 0; |
1356
|
|
|
} |
1357
|
|
|
return $ar; |
1358
|
|
|
} |
1359
|
|
|
} else { |
1360
|
|
|
for ($i = 0; $i < $this->ndim; ++$i) { |
1361
|
|
|
$ar->data[$i] = $this->data[$i] < $m ? 1 : 0; |
1362
|
|
|
} |
1363
|
|
|
return $ar; |
1364
|
|
|
} |
1365
|
|
|
} |
1366
|
|
|
|
1367
|
|
|
/** |
1368
|
|
|
* Is the matrix symmetric i.e. is it equal to its own transpose? |
1369
|
|
|
* |
1370
|
|
|
* @return bool |
1371
|
|
|
*/ |
1372
|
|
|
public function isSymmetric(): bool { |
1373
|
|
|
if (!$this->isSquare()) { |
1374
|
|
|
return false; |
1375
|
|
|
} |
1376
|
|
|
$ar = $this->transpose(); |
1377
|
|
|
for ($i = 0; $i < $ar->ndim; ++$i) { |
1378
|
|
|
if ($ar->data[$i] != $this->data[$i]) { |
1379
|
|
|
unset($ar); |
1380
|
|
|
return false; |
1381
|
|
|
} |
1382
|
|
|
} |
1383
|
|
|
unset($ar); |
1384
|
|
|
return true; |
1385
|
|
|
} |
1386
|
|
|
|
1387
|
|
|
/** |
1388
|
|
|
* print the matrix in consol |
1389
|
|
|
*/ |
1390
|
|
|
public function printMatrix() { |
1391
|
|
|
echo __CLASS__ . PHP_EOL; |
1392
|
|
|
for ($i = 0; $i < $this->row; ++$i) { |
1393
|
|
|
for ($j = 0; $j < $this->col; ++$j) { |
1394
|
|
|
printf('%lf ', $this->data[$i * $this->col + $j]); |
1395
|
|
|
} |
1396
|
|
|
echo PHP_EOL; |
1397
|
|
|
} |
1398
|
|
|
} |
1399
|
|
|
|
1400
|
|
|
public function __toString() { |
1401
|
|
|
return (string) $this->printMatrix(); |
1402
|
|
|
} |
1403
|
|
|
|
1404
|
|
|
private function flattenArray(array $ar) { |
1405
|
|
|
if (is_array($ar) && is_array($ar[0])) { |
1406
|
|
|
$a = []; |
1407
|
|
|
foreach ($ar as $y => $value) { |
1408
|
|
|
foreach ($value as $k => $v) { |
1409
|
|
|
$a[] = $v; |
1410
|
|
|
} |
1411
|
|
|
} |
1412
|
|
|
return $a; |
1413
|
|
|
} |
1414
|
|
|
} |
1415
|
|
|
|
1416
|
|
|
/** |
1417
|
|
|
* |
1418
|
|
|
* @param int $row |
1419
|
|
|
* @param int $col |
1420
|
|
|
* @param int $dtype |
1421
|
|
|
* @return $this |
1422
|
|
|
*/ |
1423
|
|
|
protected function __construct(public int $row, public int $col, int $dtype = self::Float) { |
1424
|
|
|
if ($this->row < 1 || $this->col < 1) { |
1425
|
|
|
self::_invalidArgument('* To create Numphp/Matrix row & col must be greater than 0!, Op Failed! * '); |
1426
|
|
|
} |
1427
|
|
|
parent::__construct($this->row * $this->col, $dtype); |
1428
|
|
|
return $this; |
1429
|
|
|
} |
1430
|
|
|
} |
1431
|
|
|
|