start to implement SVM with libsvm

src/Phpml/Classification/SVC.php

@@ -0,0 +1,53 @@
+<?php
+
+declare (strict_types = 1);
+
+namespace Phpml\Classification;
+
+use Phpml\Classification\Traits\Predictable;
+use Phpml\Classification\Traits\Trainable;
+use Phpml\Math\Kernel;
+
+class SVC implements Classifier
+{
+    use Trainable, Predictable;
+
+    /**
+     * @var int
+     */
+    private $kernel;
+
+    /**
+     * @var float
+     */
+    private $cost;
+
+    /**
+     * @param int   $kernel
+     * @param float $cost
+     */
+    public function __construct(int $kernel, float $cost)
+    {
+        $this->kernel = $kernel;
+        $this->cost = $cost;
+    }
+
+    /**
+     * @param array $samples
+     * @param array $labels
+     */
+    public function train(array $samples, array $labels)
+    {
+        $this->samples = $samples;
+        $this->labels = $labels;
+    }
+
+    /**
+     * @param array $sample
+     *
+     * @return mixed
+     */
+    protected function predictSample(array $sample)
+    {
+    }
+}

src/Phpml/Classification/SupportVectorMachine.php

@@ -1,78 +0,0 @@
-<?php
-
-declare (strict_types = 1);
-
-namespace Phpml\Classification;
-
-use Phpml\Classification\Traits\Predictable;
-use Phpml\Classification\Traits\Trainable;
-use Phpml\Math\Kernel;
-
-class SupportVectorMachine implements Classifier
-{
-    use Trainable, Predictable;
-
-    /**
-     * @var Kernel
-     */
-    private $kernel;
-
-    /**
-     * @var float
-     */
-    private $C;
-
-    /**
-     * @var float
-     */
-    private $tolerance;
-
-    /**
-     * @var int
-     */
-    private $upperBound;
-
-    /**
-     * @var string
-     */
-    private $binPath;
-
-    /**
-     * @param Kernel $kernel
-     * @param float  $C
-     * @param float  $tolerance
-     * @param int    $upperBound
-     */
-    public function __construct(Kernel $kernel = null, float $C = 1.0, float $tolerance = .001, int $upperBound = 100)
-    {
-        if (null === $kernel) {
-            $kernel = new Kernel\RBF($gamma = .001);
-        }
-
-        $this->kernel = $kernel;
-        $this->C = $C;
-        $this->tolerance = $tolerance;
-        $this->upperBound = $upperBound;
-
-        $this->binPath = realpath(implode(DIRECTORY_SEPARATOR, array(dirname(__FILE__), '..', '..', '..', 'bin'))) . DIRECTORY_SEPARATOR;
-    }
-
-    /**
-     * @param array $samples
-     * @param array $labels
-     */
-    public function train(array $samples, array $labels)
-    {
-        $this->samples = $samples;
-        $this->labels = $labels;
-    }
-
-    /**
-     * @param array $sample
-     *
-     * @return mixed
-     */
-    protected function predictSample(array $sample)
-    {
-    }
-}

src/Phpml/Dataset/Dataset.php

@@ -6,7 +6,6 @@ namespace Phpml\Dataset;
 
 interface Dataset
 {
-    const SOME = 'z';
     /**
      * @return array
      */

src/Phpml/SupportVectorMachine/DataTransformer.php

@@ -0,0 +1,59 @@
+<?php
+
+declare (strict_types = 1);
+
+namespace Phpml\SupportVectorMachine;
+
+class DataTransformer
+{
+    /**
+     * @param array $samples
+     * @param array $labels
+     *
+     * @return string
+     */
+    public static function trainingSet(array $samples, array $labels): string
+    {
+        $set = '';
+        $numericLabels = self::numericLabels($labels);
+        foreach ($labels as $index => $label) {
+            $set .= sprintf('%s %s %s', $numericLabels[$label], self::sampleRow($samples[$index]), PHP_EOL);
+        }
+
+        return $set;
+    }
+
+    /**
+     * @param array $labels
+     *
+     * @return array
+     */
+    public static function numericLabels(array $labels): array
+    {
+        $numericLabels = [];
+        foreach ($labels as $label) {
+            if (isset($numericLabels[$label])) {
+                continue;
+            }
+
+            $numericLabels[$label] = count($numericLabels);
+        }
+
+        return $numericLabels;
+    }
+
+    /**
+     * @param array $sample
+     *
+     * @return string
+     */
+    private static function sampleRow(array $sample): string
+    {
+        $row = [];
+        foreach ($sample as $index => $feature) {
+            $row[] = sprintf('%s:%s', $index, $feature);
+        }
+
+        return implode(' ', $row);
+    }
+}

src/Phpml/SupportVectorMachine/Kernel.php

@@ -0,0 +1,28 @@
+<?php
+
+declare (strict_types = 1);
+
+namespace Phpml\SupportVectorMachine;
+
+abstract class Kernel
+{
+    /**
+     * u'*v.
+     */
+    const LINEAR = 0;
+
+    /**
+     * (gamma*u'*v + coef0)^degree.
+     */
+    const POLYNOMIAL = 1;
+
+    /**
+     * exp(-gamma*|u-v|^2).
+     */
+    const RBF = 2;
+
+    /**
+     * tanh(gamma*u'*v + coef0).
+     */
+    const SIGMOID = 3;
+}

src/Phpml/SupportVectorMachine/SupportVectorMachine.php

@@ -0,0 +1,83 @@
+<?php
+
+declare (strict_types = 1);
+
+namespace Phpml\SupportVectorMachine;
+
+class SupportVectorMachine
+{
+    /**
+     * @var int
+     */
+    private $type;
+
+    /**
+     * @var int
+     */
+    private $kernel;
+
+    /**
+     * @var float
+     */
+    private $cost;
+
+    /**
+     * @var string
+     */
+    private $binPath;
+
+    /**
+     * @var
+     */
+    private $varPath;
+
+    /**
+     * @var string
+     */
+    private $model;
+
+    /**
+     * @param int   $type
+     * @param int   $kernel
+     * @param float $cost
+     */
+    public function __construct(int $type, int $kernel, float $cost)
+    {
+        $this->type = $type;
+        $this->kernel = $kernel;
+        $this->cost = $cost;
+
+        $rootPath = realpath(implode(DIRECTORY_SEPARATOR, [dirname(__FILE__), '..', '..', '..'])).DIRECTORY_SEPARATOR;
+
+        $this->binPath = $rootPath.'bin'.DIRECTORY_SEPARATOR.'libsvm'.DIRECTORY_SEPARATOR;
+        $this->varPath = $rootPath.'var'.DIRECTORY_SEPARATOR;
+    }
+
+    /**
+     * @param array $samples
+     * @param array $labels
+     */
+    public function train(array $samples, array $labels)
+    {
+        $trainingSet = DataTransformer::trainingSet($samples, $labels);
+        file_put_contents($trainingSetFileName = $this->varPath.uniqid(), $trainingSet);
+        $modelFileName = $trainingSetFileName.'-model';
+
+        $command = sprintf('%ssvm-train -s %s -t %s -c %s %s %s', $this->binPath, $this->type, $this->kernel, $this->cost, $trainingSetFileName, $modelFileName);
+        $output = '';
+        exec(escapeshellcmd($command), $output);
+
+        $this->model = file_get_contents($modelFileName);
+
+        unlink($trainingSetFileName);
+        unlink($modelFileName);
+    }
+
+    /**
+     * @return string
+     */
+    public function getModel()
+    {
+        return $this->model;
+    }
+}

src/Phpml/SupportVectorMachine/Type.php

@@ -0,0 +1,33 @@
+<?php
+
+declare (strict_types = 1);
+
+namespace Phpml\SupportVectorMachine;
+
+abstract class Type
+{
+    /**
+     * classification.
+     */
+    const C_SVC = 0;
+
+    /**
+     * classification.
+     */
+    const NU_SVC = 1;
+
+    /**
+     * distribution estimation.
+     */
+    const ONE_CLASS_SVM = 2;
+
+    /**
+     * regression.
+     */
+    const EPSILON_SVR = 3;
+
+    /**
+     * regression.
+     */
+    const NU_SVR = 4;
+}

tests/Phpml/SupportVectorMachine/DataTransformerTest.php

@@ -0,0 +1,25 @@
+<?php
+
+declare (strict_types = 1);
+
+namespace tests\SupportVectorMachine;
+
+use Phpml\SupportVectorMachine\DataTransformer;
+
+class DataTransformerTest extends \PHPUnit_Framework_TestCase
+{
+    public function testTransformDatasetToTrainingSet()
+    {
+        $samples = [[1, 1], [2, 1], [3, 2], [4, 5]];
+        $labels = ['a', 'a', 'b', 'b'];
+
+        $trainingSet =
+            '0 0:1 1:1 '.PHP_EOL.
+            '0 0:2 1:1 '.PHP_EOL.
+            '1 0:3 1:2 '.PHP_EOL.
+            '1 0:4 1:5 '.PHP_EOL
+        ;
+
+        $this->assertEquals($trainingSet, DataTransformer::trainingSet($samples, $labels));
+    }
+}

tests/Phpml/SupportVectorMachine/SupportVectorMachineTest.php

@@ -0,0 +1,36 @@
+<?php
+
+declare (strict_types = 1);
+
+namespace tests\SupportVectorMachine;
+
+use Phpml\SupportVectorMachine\Kernel;
+use Phpml\SupportVectorMachine\SupportVectorMachine;
+use Phpml\SupportVectorMachine\Type;
+
+class SupportVectorMachineTest extends \PHPUnit_Framework_TestCase
+{
+    public function testTrainCSVCModelWithLinearKernel()
+    {
+        $samples = [[1, 3], [1, 4], [2, 4], [3, 1], [4, 1], [4, 2]];
+        $labels = ['a', 'a', 'a', 'b', 'b', 'b'];
+
+        $model =
+            'svm_type c_svc
+kernel_type linear
+nr_class 2
+total_sv 2
+rho 0
+label 0 1
+nr_sv 1 1
+SV
+0.25 0:2 1:4 
+-0.25 0:4 1:2 
+';
+
+        $svm = new SupportVectorMachine(Type::C_SVC, Kernel::LINEAR, 100.0);
+        $svm->train($samples, $labels);
+
+        $this->assertEquals($model, $svm->getModel());
+    }
+}