Changed Backpropagation to use Java for performance reasons

Author: Zomis

src/main/groovy/net/zomis/machlearn/images/ImageNetworkBuilder.java

@@ -1,6 +1,6 @@
 package net.zomis.machlearn.images;
 
-import net.zomis.machlearn.neural.BackPropagation;
+import net.zomis.machlearn.neural.Backpropagation;
 import net.zomis.machlearn.neural.LearningData;
 import net.zomis.machlearn.neural.NeuralNetwork;
 import net.zomis.machlearn.neural.NeuronLayer;
@@ -36,7 +36,7 @@ public ImageNetworkBuilder classify(Object result, double[] input) {
         return this;
     }
 
-    public ImageNetwork learn(BackPropagation backprop, Random random) {
+    public ImageNetwork learn(Backpropagation backprop, Random random) {
         int outputNodes = classifications.size() - 1;
 
         NeuronLayer parentLayer = this.network.getLastLayer();

src/main/groovy/net/zomis/machlearn/neural/BackPropagation.groovy renamed to src/main/groovy/net/zomis/machlearn/neural/Backpropagation.java

@@ -13,4 +13,8 @@ class LearningData {
         this.outputs = outputs
     }
 
+    double getInput(int i) {
+        return inputs[i];
+    }
+
 }

src/main/groovy/net/zomis/machlearn/neural/LearningData.groovy

@@ -22,7 +22,7 @@ class NeuralMain {
         examples << new LearningData([0, 1] as double[], [0, 1] as double[])
         examples << new LearningData([1, 0] as double[], [0, 1] as double[])
         examples << new LearningData([1, 1] as double[], [1, 1] as double[])
-        new BackPropagation(0.2, 100000).backPropagationLearning(examples, network)
+        new Backpropagation(0.2, 100000).backPropagationLearning(examples, network)
 
         network.printAll()
 

src/main/groovy/net/zomis/machlearn/neural/NeuralMain.groovy

@@ -1,6 +1,6 @@
 package net.zomis.machlearn.images;
 
-import net.zomis.machlearn.neural.BackPropagation;
+import net.zomis.machlearn.neural.Backpropagation;
 import net.zomis.machlearn.neural.Neuron;
 
 import javax.imageio.ImageIO;
@@ -32,7 +32,7 @@ public static void main(String[] args) throws AWTException, IOException {
             .classify("clicked", analyze.imagePart(image, 790, 241))
             .classify("flag", analyze.imagePart(image, 790, 197))
             .classifyNone(analyze.imagePart(image, 0, 0))
-            .learn(new BackPropagation(0.1, 100), new Random(42));
+            .learn(new Backpropagation(0.1, 100), new Random(42));
 //        SlidingWindowResult points = analyze.slidingWindow(network, image).scaleX(25, 60).step(4).overlapping(false).run();
         // network.getNetwork().printAll();
 

src/test/groovy/net/zomis/machlearn/images/Screenshoter.java

@@ -57,7 +57,7 @@ class BasicPropagationTest {
 
     @Test
     void learn() {
-        new BackPropagation(0.2, 100000).backPropagationLearning(examples, network)
+        new Backpropagation(0.2, 100000).backPropagationLearning(examples, network)
 
         network.printAll()
 

src/test/groovy/net/zomis/machlearn/neural/BasicPropagationTest.groovy

@@ -25,7 +25,7 @@ class LoadSaveTest {
 
         def outputLayer = network.createLayer('OUT')
         outputLayer.createNeuron().addInputs(middleLayer)
-        new BackPropagation(0.2, 100000).backPropagationLearning(examples, network, new Random(42))
+        new Backpropagation(0.2, 100000).backPropagationLearning(examples, network, new Random(42))
 
         def savedNetwork = new ByteArrayOutputStream()
         network.save(savedNetwork)