@@ -4,7 +4,6 @@ |
||
4 | 4 | |
5 | 5 | namespace Phpml\Classification\Linear; |
6 | 6 | |
7 | -use Phpml\Classification\Classifier; |
|
8 | 7 | use Phpml\Helper\Optimizer\ConjugateGradient; |
9 | 8 | |
10 | 9 | class LogisticRegression extends Adaline |
@@ -1,6 +1,6 @@ discard block |
||
1 | 1 | <?php |
2 | 2 | |
3 | -declare(strict_types=1); |
|
3 | +declare(strict_types = 1); |
|
4 | 4 | |
5 | 5 | namespace Phpml\Classification\Linear; |
6 | 6 | |
@@ -13,12 +13,12 @@ discard block |
||
13 | 13 | /** |
14 | 14 | * Batch training: Gradient descent algorithm (default) |
15 | 15 | */ |
16 | - const BATCH_TRAINING = 1; |
|
16 | + const BATCH_TRAINING = 1; |
|
17 | 17 | |
18 | 18 | /** |
19 | 19 | * Online training: Stochastic gradient descent learning |
20 | 20 | */ |
21 | - const ONLINE_TRAINING = 2; |
|
21 | + const ONLINE_TRAINING = 2; |
|
22 | 22 | |
23 | 23 | /** |
24 | 24 | * Conjugate Batch: Conjugate Gradient algorithm |
@@ -74,14 +74,14 @@ discard block |
||
74 | 74 | string $penalty = 'L2') |
75 | 75 | { |
76 | 76 | $trainingTypes = range(self::BATCH_TRAINING, self::CONJUGATE_GRAD_TRAINING); |
77 | - if (! in_array($trainingType, $trainingTypes)) { |
|
78 | - throw new \Exception("Logistic regression can only be trained with " . |
|
79 | - "batch (gradient descent), online (stochastic gradient descent) " . |
|
77 | + if (!in_array($trainingType, $trainingTypes)) { |
|
78 | + throw new \Exception("Logistic regression can only be trained with ". |
|
79 | + "batch (gradient descent), online (stochastic gradient descent) ". |
|
80 | 80 | "or conjugate batch (conjugate gradients) algorithms"); |
81 | 81 | } |
82 | 82 | |
83 | - if (! in_array($cost, ['log', 'sse'])) { |
|
84 | - throw new \Exception("Logistic regression cost function can be one of the following: \n" . |
|
83 | + if (!in_array($cost, ['log', 'sse'])) { |
|
84 | + throw new \Exception("Logistic regression cost function can be one of the following: \n". |
|
85 | 85 | "'log' for log-likelihood and 'sse' for sum of squared errors"); |
86 | 86 | } |
87 | 87 | |
@@ -177,7 +177,7 @@ discard block |
||
177 | 177 | * The gradient of the cost function to be used with gradient descent: |
178 | 178 | * ∇J(x) = -(y - h(x)) = (h(x) - y) |
179 | 179 | */ |
180 | - $callback = function ($weights, $sample, $y) use ($penalty) { |
|
180 | + $callback = function($weights, $sample, $y) use ($penalty) { |
|
181 | 181 | $this->weights = $weights; |
182 | 182 | $hX = $this->output($sample); |
183 | 183 | |
@@ -208,7 +208,7 @@ discard block |
||
208 | 208 | * The gradient of the cost function: |
209 | 209 | * ∇J(x) = -(h(x) - y) . h(x) . (1 - h(x)) |
210 | 210 | */ |
211 | - $callback = function ($weights, $sample, $y) use ($penalty) { |
|
211 | + $callback = function($weights, $sample, $y) use ($penalty) { |
|
212 | 212 | $this->weights = $weights; |
213 | 213 | $hX = $this->output($sample); |
214 | 214 |
@@ -15,7 +15,7 @@ discard block |
||
15 | 15 | { |
16 | 16 | use Predictable, OneVsRest; |
17 | 17 | |
18 | - /** |
|
18 | + /** |
|
19 | 19 | * @var array |
20 | 20 | */ |
21 | 21 | protected $samples = []; |
@@ -83,7 +83,7 @@ discard block |
||
83 | 83 | $this->maxIterations = $maxIterations; |
84 | 84 | } |
85 | 85 | |
86 | - /** |
|
86 | + /** |
|
87 | 87 | * @param array $samples |
88 | 88 | * @param array $targets |
89 | 89 | */ |
@@ -1,6 +1,6 @@ discard block |
||
1 | 1 | <?php |
2 | 2 | |
3 | -declare(strict_types=1); |
|
3 | +declare(strict_types = 1); |
|
4 | 4 | |
5 | 5 | namespace Phpml\Classification\Linear; |
6 | 6 | |
@@ -118,7 +118,7 @@ discard block |
||
118 | 118 | protected function runTraining() |
119 | 119 | { |
120 | 120 | // The cost function is the sum of squares |
121 | - $callback = function ($weights, $sample, $target) { |
|
121 | + $callback = function($weights, $sample, $target) { |
|
122 | 122 | $this->weights = $weights; |
123 | 123 | |
124 | 124 | $prediction = $this->outputClass($sample); |
@@ -137,7 +137,7 @@ discard block |
||
137 | 137 | */ |
138 | 138 | protected function runGradientDescent(\Closure $gradientFunc, bool $isBatch = false) |
139 | 139 | { |
140 | - $class = $isBatch ? GD::class : StochasticGD::class; |
|
140 | + $class = $isBatch ? GD::class : StochasticGD::class; |
|
141 | 141 | |
142 | 142 | $optimizer = (new $class($this->featureCount)) |
143 | 143 | ->setLearningRate($this->learningRate) |
@@ -227,6 +227,6 @@ discard block |
||
227 | 227 | |
228 | 228 | $predictedClass = $this->outputClass($sample); |
229 | 229 | |
230 | - return $this->labels[ $predictedClass ]; |
|
230 | + return $this->labels[$predictedClass]; |
|
231 | 231 | } |
232 | 232 | } |
@@ -1,6 +1,6 @@ discard block |
||
1 | 1 | <?php |
2 | 2 | |
3 | -declare(strict_types=1); |
|
3 | +declare(strict_types = 1); |
|
4 | 4 | |
5 | 5 | namespace Phpml\Classification\Linear; |
6 | 6 | |
@@ -12,12 +12,12 @@ discard block |
||
12 | 12 | /** |
13 | 13 | * Batch training is the default Adaline training algorithm |
14 | 14 | */ |
15 | - const BATCH_TRAINING = 1; |
|
15 | + const BATCH_TRAINING = 1; |
|
16 | 16 | |
17 | 17 | /** |
18 | 18 | * Online training: Stochastic gradient descent learning |
19 | 19 | */ |
20 | - const ONLINE_TRAINING = 2; |
|
20 | + const ONLINE_TRAINING = 2; |
|
21 | 21 | |
22 | 22 | /** |
23 | 23 | * Training type may be either 'Batch' or 'Online' learning |
@@ -41,7 +41,7 @@ discard block |
||
41 | 41 | public function __construct(float $learningRate = 0.001, int $maxIterations = 1000, |
42 | 42 | bool $normalizeInputs = true, int $trainingType = self::BATCH_TRAINING) |
43 | 43 | { |
44 | - if (! in_array($trainingType, [self::BATCH_TRAINING, self::ONLINE_TRAINING])) { |
|
44 | + if (!in_array($trainingType, [self::BATCH_TRAINING, self::ONLINE_TRAINING])) { |
|
45 | 45 | throw new \Exception("Adaline can only be trained with batch and online/stochastic gradient descent algorithm"); |
46 | 46 | } |
47 | 47 | |
@@ -57,7 +57,7 @@ discard block |
||
57 | 57 | protected function runTraining() |
58 | 58 | { |
59 | 59 | // The cost function is the sum of squares |
60 | - $callback = function ($weights, $sample, $target) { |
|
60 | + $callback = function($weights, $sample, $target) { |
|
61 | 61 | $this->weights = $weights; |
62 | 62 | |
63 | 63 | $output = $this->output($sample); |
@@ -1,6 +1,6 @@ discard block |
||
1 | 1 | <?php |
2 | 2 | |
3 | -declare(strict_types=1); |
|
3 | +declare(strict_types = 1); |
|
4 | 4 | |
5 | 5 | namespace Phpml\Helper\Optimizer; |
6 | 6 | |
@@ -42,7 +42,7 @@ discard block |
||
42 | 42 | |
43 | 43 | $this->updateWeightsWithUpdates($updates, $totalPenalty); |
44 | 44 | |
45 | - $this->costValues[] = array_sum($errors)/$this->sampleCount; |
|
45 | + $this->costValues[] = array_sum($errors) / $this->sampleCount; |
|
46 | 46 | |
47 | 47 | if ($this->earlyStop($theta)) { |
48 | 48 | break; |
@@ -63,7 +63,7 @@ discard block |
||
63 | 63 | protected function gradient(array $theta) |
64 | 64 | { |
65 | 65 | $costs = []; |
66 | - $gradient= []; |
|
66 | + $gradient = []; |
|
67 | 67 | $totalPenalty = 0; |
68 | 68 | |
69 | 69 | foreach ($this->samples as $index => $sample) { |
@@ -73,7 +73,7 @@ discard block |
||
73 | 73 | list($cost, $grad, $penalty) = array_pad($result, 3, 0); |
74 | 74 | |
75 | 75 | $costs[] = $cost; |
76 | - $gradient[]= $grad; |
|
76 | + $gradient[] = $grad; |
|
77 | 77 | $totalPenalty += $penalty; |
78 | 78 | } |
79 | 79 | |
@@ -89,7 +89,7 @@ discard block |
||
89 | 89 | protected function updateWeightsWithUpdates(array $updates, float $penalty) |
90 | 90 | { |
91 | 91 | // Updates all weights at once |
92 | - for ($i=0; $i <= $this->dimensions; $i++) { |
|
92 | + for ($i = 0; $i <= $this->dimensions; $i++) { |
|
93 | 93 | if ($i == 0) { |
94 | 94 | $this->theta[0] -= $this->learningRate * array_sum($updates); |
95 | 95 | } else { |
@@ -1,6 +1,6 @@ discard block |
||
1 | 1 | <?php |
2 | 2 | |
3 | -declare(strict_types=1); |
|
3 | +declare(strict_types = 1); |
|
4 | 4 | |
5 | 5 | namespace Phpml\Helper\Optimizer; |
6 | 6 | |
@@ -72,7 +72,7 @@ discard block |
||
72 | 72 | * |
73 | 73 | * @var array |
74 | 74 | */ |
75 | - protected $costValues= []; |
|
75 | + protected $costValues = []; |
|
76 | 76 | |
77 | 77 | /** |
78 | 78 | * Initializes the SGD optimizer for the given number of dimensions |
@@ -216,7 +216,7 @@ discard block |
||
216 | 216 | $this->theta[0] -= $this->learningRate * $gradient; |
217 | 217 | |
218 | 218 | // Update other values |
219 | - for ($i=1; $i <= $this->dimensions; $i++) { |
|
219 | + for ($i = 1; $i <= $this->dimensions; $i++) { |
|
220 | 220 | $this->theta[$i] -= $this->learningRate * |
221 | 221 | ($gradient * $sample[$i - 1] + $penalty * $this->theta[$i]); |
222 | 222 | } |
@@ -240,7 +240,7 @@ discard block |
||
240 | 240 | { |
241 | 241 | // Check for early stop: No change larger than threshold (default 1e-5) |
242 | 242 | $diff = array_map( |
243 | - function ($w1, $w2) { |
|
243 | + function($w1, $w2) { |
|
244 | 244 | return abs($w1 - $w2) > $this->threshold ? 1 : 0; |
245 | 245 | }, |
246 | 246 | $oldTheta, $this->theta); |
@@ -1,6 +1,6 @@ discard block |
||
1 | 1 | <?php |
2 | 2 | |
3 | -declare(strict_types=1); |
|
3 | +declare(strict_types = 1); |
|
4 | 4 | |
5 | 5 | namespace Phpml\Helper\Optimizer; |
6 | 6 | |
@@ -31,7 +31,7 @@ discard block |
||
31 | 31 | |
32 | 32 | // Inits the weights randomly |
33 | 33 | $this->theta = []; |
34 | - for ($i=0; $i < $this->dimensions; $i++) { |
|
34 | + for ($i = 0; $i < $this->dimensions; $i++) { |
|
35 | 35 | $this->theta[] = rand() / (float) getrandmax(); |
36 | 36 | } |
37 | 37 | } |
@@ -1,6 +1,6 @@ discard block |
||
1 | 1 | <?php |
2 | 2 | |
3 | -declare(strict_types=1); |
|
3 | +declare(strict_types = 1); |
|
4 | 4 | |
5 | 5 | namespace Phpml\Helper\Optimizer; |
6 | 6 | |
@@ -34,7 +34,7 @@ discard block |
||
34 | 34 | |
35 | 35 | $d = mp::muls($this->gradient($this->theta), -1); |
36 | 36 | |
37 | - for ($i=0; $i < $this->maxIterations; $i++) { |
|
37 | + for ($i = 0; $i < $this->maxIterations; $i++) { |
|
38 | 38 | // Obtain α that minimizes f(θ + α.d) |
39 | 39 | $alpha = $this->getAlpha(array_sum($d)); |
40 | 40 | |
@@ -161,7 +161,7 @@ discard block |
||
161 | 161 | { |
162 | 162 | $theta = $this->theta; |
163 | 163 | |
164 | - for ($i=0; $i < $this->dimensions + 1; $i++) { |
|
164 | + for ($i = 0; $i < $this->dimensions + 1; $i++) { |
|
165 | 165 | if ($i == 0) { |
166 | 166 | $theta[$i] += $alpha * array_sum($d); |
167 | 167 | } else { |