diff --git a/docs/machine-learning/neural-network/multilayer-perceptron-classifier.md b/docs/machine-learning/neural-network/multilayer-perceptron-classifier.md
index 72d0b4b..5acf093 100644
--- a/docs/machine-learning/neural-network/multilayer-perceptron-classifier.md
+++ b/docs/machine-learning/neural-network/multilayer-perceptron-classifier.md
@@ -66,4 +66,6 @@ $mlp->predict([[1, 1, 1, 1], [0, 0, 0, 0]]);
 * BinaryStep
 * Gaussian
 * HyperbolicTangent
+* Parametric Rectified Linear Unit
 * Sigmoid (default)
+* Thresholded Rectified Linear Unit
diff --git a/src/Phpml/NeuralNetwork/ActivationFunction.php b/src/Phpml/NeuralNetwork/ActivationFunction.php
index 5b91425..30adf4d 100644
--- a/src/Phpml/NeuralNetwork/ActivationFunction.php
+++ b/src/Phpml/NeuralNetwork/ActivationFunction.php
@@ -10,4 +10,10 @@ interface ActivationFunction
      * @param float|int $value
      */
     public function compute($value): float;
+
+    /**
+     * @param float|int $value
+     * @param float|int $computedvalue
+     */
+    public function differentiate($value, $computedvalue): float;
 }
diff --git a/src/Phpml/NeuralNetwork/ActivationFunction/BinaryStep.php b/src/Phpml/NeuralNetwork/ActivationFunction/BinaryStep.php
index 764bc4e..56ea7eb 100644
--- a/src/Phpml/NeuralNetwork/ActivationFunction/BinaryStep.php
+++ b/src/Phpml/NeuralNetwork/ActivationFunction/BinaryStep.php
@@ -15,4 +15,17 @@ class BinaryStep implements ActivationFunction
     {
         return $value >= 0 ? 1.0 : 0.0;
     }
+
+    /**
+     * @param float|int $value
+     * @param float|int $computedvalue
+     */
+    public function differentiate($value, $computedvalue): float
+    {
+        if ($value === 0 || $value === 0.0) {
+            return 1;
+        }
+
+        return 0;
+    }
 }
diff --git a/src/Phpml/NeuralNetwork/ActivationFunction/Gaussian.php b/src/Phpml/NeuralNetwork/ActivationFunction/Gaussian.php
index da428a4..8871b58 100644
--- a/src/Phpml/NeuralNetwork/ActivationFunction/Gaussian.php
+++ b/src/Phpml/NeuralNetwork/ActivationFunction/Gaussian.php
@@ -15,4 +15,13 @@ class Gaussian implements ActivationFunction
     {
         return exp(-pow($value, 2));
     }
+
+    /**
+     * @param float|int $value
+     * @param float|int $calculatedvalue
+     */
+    public function differentiate($value, $calculatedvalue): float
+    {
+        return -2 * $value * $calculatedvalue;
+    }
 }
diff --git a/src/Phpml/NeuralNetwork/ActivationFunction/HyperbolicTangent.php b/src/Phpml/NeuralNetwork/ActivationFunction/HyperbolicTangent.php
index 6378606..7aa9614 100644
--- a/src/Phpml/NeuralNetwork/ActivationFunction/HyperbolicTangent.php
+++ b/src/Phpml/NeuralNetwork/ActivationFunction/HyperbolicTangent.php
@@ -25,4 +25,13 @@ class HyperbolicTangent implements ActivationFunction
     {
         return tanh($this->beta * $value);
     }
+
+    /**
+     * @param float|int $value
+     * @param float|int $computedvalue
+     */
+    public function differentiate($value, $computedvalue): float
+    {
+        return 1 - pow($computedvalue, 2);
+    }
 }
diff --git a/src/Phpml/NeuralNetwork/ActivationFunction/PReLU.php b/src/Phpml/NeuralNetwork/ActivationFunction/PReLU.php
index fc7ff62..88212d1 100644
--- a/src/Phpml/NeuralNetwork/ActivationFunction/PReLU.php
+++ b/src/Phpml/NeuralNetwork/ActivationFunction/PReLU.php
@@ -25,4 +25,13 @@ class PReLU implements ActivationFunction
     {
         return $value >= 0 ? $value : $this->beta * $value;
     }
+
+    /**
+     * @param float|int $value
+     * @param float|int $computedvalue
+     */
+    public function differentiate($value, $computedvalue): float
+    {
+        return $computedvalue >= 0 ? 1.0 : $this->beta;
+    }
 }
diff --git a/src/Phpml/NeuralNetwork/ActivationFunction/Sigmoid.php b/src/Phpml/NeuralNetwork/ActivationFunction/Sigmoid.php
index 4ae9603..edad3d6 100644
--- a/src/Phpml/NeuralNetwork/ActivationFunction/Sigmoid.php
+++ b/src/Phpml/NeuralNetwork/ActivationFunction/Sigmoid.php
@@ -25,4 +25,13 @@ class Sigmoid implements ActivationFunction
     {
         return 1 / (1 + exp(-$this->beta * $value));
     }
+
+    /**
+     * @param float|int $value
+     * @param float|int $computedvalue
+     */
+    public function differentiate($value, $computedvalue): float
+    {
+        return $computedvalue * (1 - $computedvalue);
+    }
 }
diff --git a/src/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLU.php b/src/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLU.php
index 2bb1cc7..f8f8247 100644
--- a/src/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLU.php
+++ b/src/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLU.php
@@ -13,7 +13,7 @@ class ThresholdedReLU implements ActivationFunction
      */
     private $theta;
 
-    public function __construct(float $theta = 1.0)
+    public function __construct(float $theta = 0.0)
     {
         $this->theta = $theta;
     }
@@ -25,4 +25,13 @@ class ThresholdedReLU implements ActivationFunction
     {
         return $value > $this->theta ? $value : 0.0;
     }
+
+    /**
+     * @param float|int $value
+     * @param float|int $calculatedvalue
+     */
+    public function differentiate($value, $calculatedvalue): float
+    {
+        return $calculatedvalue >= $this->theta ? 1.0 : 0.0;
+    }
 }
diff --git a/src/Phpml/NeuralNetwork/Network/MultilayerPerceptron.php b/src/Phpml/NeuralNetwork/Network/MultilayerPerceptron.php
index 1a997be..bfec929 100644
--- a/src/Phpml/NeuralNetwork/Network/MultilayerPerceptron.php
+++ b/src/Phpml/NeuralNetwork/Network/MultilayerPerceptron.php
@@ -9,6 +9,7 @@ use Phpml\Exception\InvalidArgumentException;
 use Phpml\Helper\Predictable;
 use Phpml\IncrementalEstimator;
 use Phpml\NeuralNetwork\ActivationFunction;
+use Phpml\NeuralNetwork\ActivationFunction\Sigmoid;
 use Phpml\NeuralNetwork\Layer;
 use Phpml\NeuralNetwork\Node\Bias;
 use Phpml\NeuralNetwork\Node\Input;
@@ -125,7 +126,10 @@ abstract class MultilayerPerceptron extends LayeredNetwork implements Estimator,
     {
         $this->addInputLayer($this->inputLayerFeatures);
         $this->addNeuronLayers($this->hiddenLayers, $this->activationFunction);
-        $this->addNeuronLayers([count($this->classes)], $this->activationFunction);
+
+        // Sigmoid function for the output layer as we want a value from 0 to 1.
+        $sigmoid = new Sigmoid();
+        $this->addNeuronLayers([count($this->classes)], $sigmoid);
 
         $this->addBiasNodes();
         $this->generateSynapses();
diff --git a/src/Phpml/NeuralNetwork/Node/Neuron.php b/src/Phpml/NeuralNetwork/Node/Neuron.php
index 2dff600..47d606d 100644
--- a/src/Phpml/NeuralNetwork/Node/Neuron.php
+++ b/src/Phpml/NeuralNetwork/Node/Neuron.php
@@ -26,6 +26,11 @@ class Neuron implements Node
      */
     protected $output = 0.0;
 
+    /**
+     * @var float
+     */
+    protected $z = 0.0;
+
     public function __construct(?ActivationFunction $activationFunction = null)
     {
         $this->activationFunction = $activationFunction ?: new Sigmoid();
@@ -47,19 +52,25 @@ class Neuron implements Node
     public function getOutput(): float
     {
         if ($this->output === 0.0) {
-            $sum = 0.0;
+            $this->z = 0;
             foreach ($this->synapses as $synapse) {
-                $sum += $synapse->getOutput();
+                $this->z += $synapse->getOutput();
             }
 
-            $this->output = $this->activationFunction->compute($sum);
+            $this->output = $this->activationFunction->compute($this->z);
         }
 
         return $this->output;
     }
 
+    public function getDerivative(): float
+    {
+        return $this->activationFunction->differentiate($this->z, $this->output);
+    }
+
     public function reset(): void
     {
         $this->output = 0.0;
+        $this->z = 0.0;
     }
 }
diff --git a/src/Phpml/NeuralNetwork/Training/Backpropagation.php b/src/Phpml/NeuralNetwork/Training/Backpropagation.php
index fd09d95..6c9af98 100644
--- a/src/Phpml/NeuralNetwork/Training/Backpropagation.php
+++ b/src/Phpml/NeuralNetwork/Training/Backpropagation.php
@@ -64,7 +64,7 @@ class Backpropagation
     private function getSigma(Neuron $neuron, int $targetClass, int $key, bool $lastLayer): float
     {
         $neuronOutput = $neuron->getOutput();
-        $sigma = $neuronOutput * (1 - $neuronOutput);
+        $sigma = $neuron->getDerivative();
 
         if ($lastLayer) {
             $value = 0;
diff --git a/tests/Phpml/Classification/MLPClassifierTest.php b/tests/Phpml/Classification/MLPClassifierTest.php
index ef40618..c46e297 100644
--- a/tests/Phpml/Classification/MLPClassifierTest.php
+++ b/tests/Phpml/Classification/MLPClassifierTest.php
@@ -7,6 +7,11 @@ namespace Phpml\Tests\Classification;
 use Phpml\Classification\MLPClassifier;
 use Phpml\Exception\InvalidArgumentException;
 use Phpml\ModelManager;
+use Phpml\NeuralNetwork\ActivationFunction;
+use Phpml\NeuralNetwork\ActivationFunction\HyperbolicTangent;
+use Phpml\NeuralNetwork\ActivationFunction\PReLU;
+use Phpml\NeuralNetwork\ActivationFunction\Sigmoid;
+use Phpml\NeuralNetwork\ActivationFunction\ThresholdedReLU;
 use Phpml\NeuralNetwork\Node\Neuron;
 use PHPUnit\Framework\TestCase;
 
@@ -141,6 +146,33 @@ class MLPClassifierTest extends TestCase
         $this->assertEquals(4, $network->predict([0, 0, 0, 0, 0]));
     }
 
+    /**
+     * @dataProvider activationFunctionsProvider
+     */
+    public function testBackpropagationActivationFunctions(ActivationFunction $activationFunction): void
+    {
+        $network = new MLPClassifier(5, [3], ['a', 'b'], 10000, $activationFunction);
+        $network->train(
+            [[1, 0, 0, 0, 0], [0, 1, 0, 0, 0], [0, 0, 1, 1, 0], [1, 1, 1, 1, 1]],
+            ['a', 'b', 'a', 'a']
+        );
+
+        $this->assertEquals('a', $network->predict([1, 0, 0, 0, 0]));
+        $this->assertEquals('b', $network->predict([0, 1, 0, 0, 0]));
+        $this->assertEquals('a', $network->predict([0, 0, 1, 1, 0]));
+        $this->assertEquals('a', $network->predict([1, 1, 1, 1, 1]));
+    }
+
+    public function activationFunctionsProvider(): array
+    {
+        return [
+            [new Sigmoid()],
+            [new HyperbolicTangent()],
+            [new PReLU()],
+            [new ThresholdedReLU()],
+        ];
+    }
+
     public function testSaveAndRestore(): void
     {
         // Instantinate new Percetron trained for OR problem
diff --git a/tests/Phpml/NeuralNetwork/ActivationFunction/BinaryStepTest.php b/tests/Phpml/NeuralNetwork/ActivationFunction/BinaryStepTest.php
index acc7977..4e85478 100644
--- a/tests/Phpml/NeuralNetwork/ActivationFunction/BinaryStepTest.php
+++ b/tests/Phpml/NeuralNetwork/ActivationFunction/BinaryStepTest.php
@@ -27,4 +27,23 @@ class BinaryStepTest extends TestCase
             [0, -0.1],
         ];
     }
+
+    /**
+     * @dataProvider binaryStepDerivativeProvider
+     */
+    public function testBinaryStepDerivative($expected, $value): void
+    {
+        $binaryStep = new BinaryStep();
+        $activatedValue = $binaryStep->compute($value);
+        $this->assertEquals($expected, $binaryStep->differentiate($value, $activatedValue));
+    }
+
+    public function binaryStepDerivativeProvider(): array
+    {
+        return [
+            [0, -1],
+            [1, 0],
+            [0, 1],
+        ];
+    }
 }
diff --git a/tests/Phpml/NeuralNetwork/ActivationFunction/GaussianTest.php b/tests/Phpml/NeuralNetwork/ActivationFunction/GaussianTest.php
index 2b08793..aace8bc 100644
--- a/tests/Phpml/NeuralNetwork/ActivationFunction/GaussianTest.php
+++ b/tests/Phpml/NeuralNetwork/ActivationFunction/GaussianTest.php
@@ -29,4 +29,27 @@ class GaussianTest extends TestCase
             [0, -3],
         ];
     }
+
+    /**
+     * @dataProvider gaussianDerivativeProvider
+     */
+    public function testGaussianDerivative($expected, $value): void
+    {
+        $gaussian = new Gaussian();
+        $activatedValue = $gaussian->compute($value);
+        $this->assertEquals($expected, $gaussian->differentiate($value, $activatedValue), '', 0.001);
+    }
+
+    public function gaussianDerivativeProvider(): array
+    {
+        return [
+            [0, -5],
+            [0.735, -1],
+            [0.779, -0.5],
+            [0, 0],
+            [-0.779, 0.5],
+            [-0.735, 1],
+            [0, 5],
+        ];
+    }
 }
diff --git a/tests/Phpml/NeuralNetwork/ActivationFunction/HyperboliTangentTest.php b/tests/Phpml/NeuralNetwork/ActivationFunction/HyperboliTangentTest.php
index 91e7eba..629200e 100644
--- a/tests/Phpml/NeuralNetwork/ActivationFunction/HyperboliTangentTest.php
+++ b/tests/Phpml/NeuralNetwork/ActivationFunction/HyperboliTangentTest.php
@@ -30,4 +30,28 @@ class HyperboliTangentTest extends TestCase
             [0.3, 0, 0],
         ];
     }
+
+    /**
+     * @dataProvider tanhDerivativeProvider
+     */
+    public function testHyperbolicTangentDerivative($beta, $expected, $value): void
+    {
+        $tanh = new HyperbolicTangent($beta);
+        $activatedValue = $tanh->compute($value);
+        $this->assertEquals($expected, $tanh->differentiate($value, $activatedValue), '', 0.001);
+    }
+
+    public function tanhDerivativeProvider(): array
+    {
+        return [
+            [1.0, 0, -6],
+            [1.0, 0.419, -1],
+            [1.0, 1, 0],
+            [1.0, 0.419, 1],
+            [1.0, 0, 6],
+            [0.5, 0.786, 1],
+            [0.5, 0.786, -1],
+            [0.3, 1, 0],
+        ];
+    }
 }
diff --git a/tests/Phpml/NeuralNetwork/ActivationFunction/PReLUTest.php b/tests/Phpml/NeuralNetwork/ActivationFunction/PReLUTest.php
index f407060..c9f565d 100644
--- a/tests/Phpml/NeuralNetwork/ActivationFunction/PReLUTest.php
+++ b/tests/Phpml/NeuralNetwork/ActivationFunction/PReLUTest.php
@@ -29,4 +29,27 @@ class PReLUTest extends TestCase
             [0.02, -0.06, -3],
         ];
     }
+
+    /**
+     * @dataProvider preluDerivativeProvider
+     */
+    public function testPReLUDerivative($beta, $expected, $value): void
+    {
+        $prelu = new PReLU($beta);
+        $activatedValue = $prelu->compute($value);
+        $this->assertEquals($expected, $prelu->differentiate($value, $activatedValue));
+    }
+
+    public function preluDerivativeProvider(): array
+    {
+        return [
+            [0.5, 0.5, -3],
+            [0.5, 1, 0],
+            [0.5, 1, 1],
+            [0.01, 1, 1],
+            [1, 1, 1],
+            [0.3, 1, 0.1],
+            [0.1, 0.1, -0.1],
+        ];
+    }
 }
diff --git a/tests/Phpml/NeuralNetwork/ActivationFunction/SigmoidTest.php b/tests/Phpml/NeuralNetwork/ActivationFunction/SigmoidTest.php
index 30b50f8..1028fb3 100644
--- a/tests/Phpml/NeuralNetwork/ActivationFunction/SigmoidTest.php
+++ b/tests/Phpml/NeuralNetwork/ActivationFunction/SigmoidTest.php
@@ -30,4 +30,28 @@ class SigmoidTest extends TestCase
             [2.0, 0, -3.75],
         ];
     }
+
+    /**
+     * @dataProvider sigmoidDerivativeProvider
+     */
+    public function testSigmoidDerivative($beta, $expected, $value): void
+    {
+        $sigmoid = new Sigmoid($beta);
+        $activatedValue = $sigmoid->compute($value);
+        $this->assertEquals($expected, $sigmoid->differentiate($value, $activatedValue), '', 0.001);
+    }
+
+    public function sigmoidDerivativeProvider(): array
+    {
+        return [
+            [1.0, 0, -10],
+            [1, 0.006, -5],
+            [1.0, 0.25, 0],
+            [1, 0.006, 5],
+            [1.0, 0, 10],
+            [2.0, 0.25, 0],
+            [0.5, 0.246, 0.5],
+            [0.5, 0.241, 0.75],
+        ];
+    }
 }
diff --git a/tests/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLUTest.php b/tests/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLUTest.php
index f46ff02..4db0418 100644
--- a/tests/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLUTest.php
+++ b/tests/Phpml/NeuralNetwork/ActivationFunction/ThresholdedReLUTest.php
@@ -28,4 +28,26 @@ class ThresholdedReLUTest extends TestCase
             [0.9, 0, 0.1],
         ];
     }
+
+    /**
+     * @dataProvider thresholdDerivativeProvider
+     */
+    public function testThresholdedReLUDerivative($theta, $expected, $value): void
+    {
+        $thresholdedReLU = new ThresholdedReLU($theta);
+        $activatedValue = $thresholdedReLU->compute($value);
+        $this->assertEquals($expected, $thresholdedReLU->differentiate($value, $activatedValue));
+    }
+
+    public function thresholdDerivativeProvider(): array
+    {
+        return [
+            [0, 1, 1],
+            [0, 1, 0],
+            [0.5, 1, 1],
+            [0.5, 1, 1],
+            [0.5, 0, 0],
+            [2, 0, -1],
+        ];
+    }
 }