Fully fixed gradient descent

src/main/java/com/naaturel/ANN/Main.java

@@ -26,6 +26,13 @@ public class Main {
                 Map.entry(new DataSetEntry(List.of(1.0F, 1.0F)), new Label(1.0F))
         ));
 
+        DataSet andDataSet = new DataSet(Map.ofEntries(
+                Map.entry(new DataSetEntry(List.of(0.0F, 0.0F)), new Label(-1.0F)),
+                Map.entry(new DataSetEntry(List.of(0.0F, 1.0F)), new Label(-1.0F)),
+                Map.entry(new DataSetEntry(List.of(1.0F, 0.0F)), new Label(-1.0F)),
+                Map.entry(new DataSetEntry(List.of(1.0F, 1.0F)), new Label(1.0F))
+        ));
+
         DataSet dataSet = new DataSet(Map.ofEntries(
                 Map.entry(new DataSetEntry(List.of(1.0F, 6.0F)), new Label(1.0F)),
                 Map.entry(new DataSetEntry(List.of(7.0F, 9.0F)), new Label(-1.0F)),
@@ -51,17 +58,17 @@ public class Main {
         ));
 
         List<Synapse> syns = new ArrayList<>();
-        syns.add(new Synapse(new Input(0), new Weight()));
+        syns.add(new Synapse(new Input(0), new Weight(0)));
-        syns.add(new Synapse(new Input(0), new Weight()));
+        syns.add(new Synapse(new Input(0), new Weight(0)));
 
-        Bias bias =  new Bias(new Weight());
+        Bias bias =  new Bias(new Weight(0));
 
         Neuron n = new SimplePerceptron(syns, bias, new Linear());
         GradientDescentTraining st = new GradientDescentTraining();
 
         long start = System.currentTimeMillis();
 
-        st.train(n, 0.03F, orDataSet.toNormalized());
+        st.train(n, 0.2F, andDataSet);
 
         long end = System.currentTimeMillis();
         System.out.printf("Training completed in %.2f s%n", (end - start) / 1000.0);

src/main/java/com/naaturel/ANN/domain/model/dataset/DataSet.java

@@ -14,6 +14,10 @@ public class DataSet implements Iterable<DataSetEntry>{
         this.data = data;
     }
 
+    public int size() {
+        return data.size();
+    }
+
     public List<DataSetEntry> getData(){
         return new ArrayList<>(this.data.keySet());
     }

src/main/java/com/naaturel/ANN/implementation/training/GradientDescentTraining.java

@@ -19,7 +19,7 @@ public class GradientDescentTraining {
 
     public void train(Neuron n, float learningRate, DataSet dataSet) {
         int epoch = 1;
-        int maxEpoch = 10000;
+        int maxEpoch = 200;
         float errorThreshold = 0.125F;
         float currentError;
 
@@ -37,30 +37,28 @@ public class GradientDescentTraining {
                 float delta = this.calculateDelta(expectation, prediction);
                 float loss = this.calculateLoss(delta);
 
-                currentError += loss;
+                currentError += loss/dataSet.size();
 
-                Bias b = n.getBias();
+                biasCorrector += learningRate * delta * n.getBias().getInput();
-                biasCorrector += this.calculateWeightCorrection(learningRate, b.getInput(), delta);
 
                 for(int i = 0; i < correctorTerms.size(); i++){
                     Synapse syn = n.getSynapse(i);
                     float c = correctorTerms.get(i);
-                    c += this.calculateWeightCorrection(learningRate, syn.getInput(), delta);
+                    c += learningRate * delta * syn.getInput();
                     correctorTerms.set(i, c);
                 }
 
-                if(epoch % 10 != 0) continue;
                 System.out.printf("Epoch : %d ", epoch);
                 System.out.printf("predicted : %.2f, ", prediction);
                 System.out.printf("expected : %.2f, ", expectation);
                 System.out.printf("delta : %.2f, ", delta);
                 System.out.printf("loss : %.2f\n", loss);
-
             }
-
             System.out.printf("[Total error : %.3f]\n", currentError);
-            biasCorrector += n.getBias().getWeight();
+
-            n.updateBias(new Weight(biasCorrector));
+            float currentBias = n.getBias().getWeight();
+            float newBias = currentBias + biasCorrector;
+            n.updateBias(new Weight(newBias));
 
             for(int i = 0; i < correctorTerms.size(); i++){
                 Synapse syn = n.getSynapse(i);
@@ -97,8 +95,8 @@ public class GradientDescentTraining {
         return ((float) Math.pow(delta, 2))/2;
     }
 
-    private float calculateWeightCorrection(float lr, float value, float delta){
+    private float calculateWeightCorrection(float value, float delta){
-        return lr * value * delta;
+        return value * delta;
     }
 
 }