Async and parallel weights and biases changes

xEcho1337 · Oct 22, 2024 · 6626b97 · 6626b97
1 parent b420cfb
commit 6626b97
Show file tree

Hide file tree

Showing 3 changed files with 18 additions and 16 deletions.
diff --git a/src/main/java/net/echo/brain4j/structure/Neuron.java b/src/main/java/net/echo/brain4j/structure/Neuron.java
@@ -7,8 +7,7 @@ public class Neuron {
     private Synapse synapse;
     private double delta;
     private double value;
-    @Expose
-    private double bias = 2 * Math.random() - 1;
+    @Expose private double bias = 2 * Math.random() - 1;
 
     public Synapse getSynapse() {
         return synapse;

diff --git a/src/main/java/net/echo/brain4j/training/BackPropagation.java b/src/main/java/net/echo/brain4j/training/BackPropagation.java
@@ -2,15 +2,14 @@
 
 import net.echo.brain4j.layer.Layer;
 import net.echo.brain4j.layer.impl.DropoutLayer;
-import net.echo.brain4j.loss.LossFunction;
 import net.echo.brain4j.model.Model;
 import net.echo.brain4j.structure.Neuron;
 import net.echo.brain4j.structure.Synapse;
 import net.echo.brain4j.training.data.DataRow;
 import net.echo.brain4j.training.data.DataSet;
 import net.echo.brain4j.training.optimizers.Optimizer;
 
-import java.util.List;
+import java.util.*;
 
 public class BackPropagation {
 
@@ -69,6 +68,7 @@ private void initialDelta(List<Layer> layers, double[] targets, double[] outputs
 
         for (int i = 0; i < outputLayer.getNeurons().size(); i++) {
             Neuron neuron = outputLayer.getNeuronAt(i);
+
             double output = outputs[i];
             double error = targets[i] - output;
 
@@ -80,9 +80,7 @@ private void initialDelta(List<Layer> layers, double[] targets, double[] outputs
     private void updateWeightsAndBiases(List<Layer> layers, double learningRate) {
         timestep++;
 
-        for (int l = 0; l < layers.size() - 1; l++) {
-            Layer nextLayer = layers.get(l + 1);
-
+        layers.parallelStream().forEach(nextLayer -> {
             for (Synapse synapse : nextLayer.getSynapses()) {
                 optimizer.update(synapse, timestep);
             }
@@ -91,6 +89,6 @@ private void updateWeightsAndBiases(List<Layer> layers, double learningRate) {
                 double deltaBias = learningRate * neuron.getDelta();
                 neuron.setBias(neuron.getBias() + deltaBias);
             }
-        }
+        });
     }
 }
diff --git a/src/test/java/XorTest.java b/src/test/java/XorTest.java
@@ -13,13 +13,13 @@
 public class XorTest {
 
     public static void main(String[] args) {
-        Model network = new FeedForwardModel(
-                new DenseLayer(2, Activations.LINEAR),
-                new DenseLayer(32, Activations.RELU),
-                new DenseLayer(32, Activations.RELU),
-                new DenseLayer(32, Activations.RELU),
-                new DenseLayer(1, Activations.SIGMOID)
-        );
+        Model network = new FeedForwardModel();
+
+        network.add(new DenseLayer(2, Activations.LINEAR));
+        network.add(new DenseLayer(32, Activations.RELU));
+        network.add(new DenseLayer(32, Activations.RELU));
+        network.add(new DenseLayer(32, Activations.RELU));
+        network.add(new DenseLayer(1, Activations.SIGMOID));
 
         network.compile(InitializationType.XAVIER, LossFunctions.BINARY_CROSS_ENTROPY, new Adam(0.001));
 
@@ -41,12 +41,17 @@ public static void main(String[] args) {
             epoches++;
 
             network.fit(training);
+
+            double evalStart = System.nanoTime();
             error = network.evaluate(training);
+            double evalTook = System.nanoTime() - evalStart;
 
             if (epoches % 100 == 0) {
+
                 System.out.println("Epoch #" + epoches + " has error " + error);
+                System.out.println("Eval took " + (evalTook / 1e6) + "ms");
             }
-        } while (error > 0.01);
+        } while (error > 1.0E-4);
 
         double took = (System.nanoTime() - start) / 1e6;