Naaturel/ANN-framework
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Initial commit

Browse Source
This commit is contained in:
Laurent
2026-03-17 21:28:37 +01:00
commit efab887521
28 changed files with 969 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

57
src/main/java/com/naaturel/ANN/Main.java Normal file
View File

@@ -0,0 +1,57 @@
package com.naaturel.ANN;
import com.naaturel.ANN.domain.abstraction.Neuron;
import com.naaturel.ANN.domain.model.dataset.DataSet;
import com.naaturel.ANN.domain.model.dataset.DataSetEntry;
import com.naaturel.ANN.domain.model.dataset.Label;
import com.naaturel.ANN.domain.model.neuron.Bias;
import com.naaturel.ANN.domain.model.neuron.Input;
import com.naaturel.ANN.domain.model.neuron.Synapse;
import com.naaturel.ANN.domain.model.neuron.Weight;
import com.naaturel.ANN.implementation.activationFunction.Heaviside;
import com.naaturel.ANN.implementation.activationFunction.Linear;
import com.naaturel.ANN.implementation.neuron.SimplePerceptron;
import com.naaturel.ANN.implementation.training.GradientDescentTraining;
import java.util.*;
public class Main {
public static void main(String[] args){
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)),
Map.entry(new DataSetEntry(List.of(1.0F, 9.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(7.0F, 10.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(2.0F, 5.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(2.0F, 7.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(2.0F, 8.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(6.0F, 8.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(6.0F, 9.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(3.0F, 5.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(3.0F, 6.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(3.0F, 8.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(3.0F, 9.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(5.0F, 7.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(5.0F, 8.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(5.0F, 10.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(5.0F, 11.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(4.0F, 6.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(4.0F, 7.0F)), new Label(-1.0F)),
Map.entry(new DataSetEntry(List.of(4.0F, 9.0F)), new Label(1.0F)),
Map.entry(new DataSetEntry(List.of(4.0F, 10.0F)), new Label(1.0F))
));
List<Synapse> syns = new ArrayList<>();
syns.add(new Synapse(new Input(0), new Weight()));
syns.add(new Synapse(new Input(0), new Weight()));
Bias bias = new Bias(new Weight());
Neuron n = new SimplePerceptron(syns, bias, new Linear());
GradientDescentTraining st = new GradientDescentTraining();
st.train(n, 0.0003F, dataSet);
}
}

7
src/main/java/com/naaturel/ANN/domain/abstraction/ActivationFunction.java Normal file
View File

@@ -0,0 +1,7 @@
package com.naaturel.ANN.domain.abstraction;
public interface ActivationFunction {
float accept(Neuron n);
}

55
src/main/java/com/naaturel/ANN/domain/abstraction/Neuron.java Normal file
View File

@@ -0,0 +1,55 @@
package com.naaturel.ANN.domain.abstraction;
import com.naaturel.ANN.domain.model.neuron.Bias;
import com.naaturel.ANN.domain.model.neuron.Input;
import com.naaturel.ANN.domain.model.neuron.Synapse;
import com.naaturel.ANN.domain.model.neuron.Weight;
import java.util.ArrayList;
import java.util.List;
public abstract class Neuron {
protected List<Synapse> synapses;
protected Bias bias;
protected ActivationFunction activationFunction;
public Neuron(List<Synapse> synapses, Bias bias, ActivationFunction func){
this.synapses = synapses;
this.bias = bias;
this.activationFunction = func;
}
public abstract float predict();
public abstract float calculateWeightedSum();
public int getSynCount(){
return this.synapses.size();
}
public void setInput(int index, Input input){
Synapse syn = this.synapses.get(index);
syn.setInput(input.getValue());
}
public Bias getBias(){
return this.bias;
}
public void updateBias(Weight weight) {
this.bias.setWeight(weight.getValue());
}
public Synapse getSynapse(int index){
return this.synapses.get(index);
}
public List<Synapse> getSynapses() {
return new ArrayList<>(this.synapses);
}
public void setWeight(int index, Weight weight){
Synapse syn = this.synapses.get(index);
syn.setWeight(weight.getValue());
}
}

13
src/main/java/com/naaturel/ANN/domain/abstraction/NeuronTrainer.java Normal file
View File

@@ -0,0 +1,13 @@
package com.naaturel.ANN.domain.abstraction;
public abstract class NeuronTrainer {
private Trainable trainable;
public NeuronTrainer(Trainable trainable){
this.trainable = trainable;
}
public abstract void train();
}

7
src/main/java/com/naaturel/ANN/domain/abstraction/Trainable.java Normal file
View File

@@ -0,0 +1,7 @@
package com.naaturel.ANN.domain.abstraction;
public interface Trainable {
}

30
src/main/java/com/naaturel/ANN/domain/model/dataset/DataSet.java Normal file
View File

@@ -0,0 +1,30 @@
package com.naaturel.ANN.domain.model.dataset;
import java.util.*;
public class DataSet implements Iterable<DataSetEntry>{
private Map<DataSetEntry, Label> data;
public DataSet(){
this(new HashMap<>());
}
public DataSet(Map<DataSetEntry, Label> data){
this.data = data;
}
public List<DataSetEntry> getData(){
return new ArrayList<>(this.data.keySet());
}
public Label getLabel(DataSetEntry entry){
return this.data.get(entry);
}
@Override
public Iterator<DataSetEntry> iterator() {
return this.data.keySet().iterator();
}
}

40
src/main/java/com/naaturel/ANN/domain/model/dataset/DataSetEntry.java Normal file
View File

@@ -0,0 +1,40 @@
package com.naaturel.ANN.domain.model.dataset;
import java.util.*;
public class DataSetEntry implements Iterable<Float> {
private List<Float> data;
public DataSetEntry(List<Float> data){
this.data = data;
}
public List<Float> getData() {
return new ArrayList<>(data);
}
@Override
public int hashCode() {
return Objects.hash(this.data);
}
@Override
public boolean equals(Object obj) {
if (this == obj) return true;
if (!(obj instanceof DataSetEntry dataSetEntry)) return false;
return Objects.equals(this.data, dataSetEntry.data);
}
@Override
public Iterator<Float> iterator() {
return this.data.iterator();
}
@Override
public String toString() {
return Arrays.toString(this.data.toArray());
}
}

15
src/main/java/com/naaturel/ANN/domain/model/dataset/Label.java Normal file
View File

@@ -0,0 +1,15 @@
package com.naaturel.ANN.domain.model.dataset;
public class Label {
private float value;
public Label(float value){
this.value = value;
}
public float getValue() {
return value;
}
}

8
src/main/java/com/naaturel/ANN/domain/model/neuron/Bias.java Normal file
View File

@@ -0,0 +1,8 @@
package com.naaturel.ANN.domain.model.neuron;
public class Bias extends Synapse {
public Bias(Weight weight) {
super(new Input(1), weight);
}
}

20
src/main/java/com/naaturel/ANN/domain/model/neuron/Input.java Normal file
View File

@@ -0,0 +1,20 @@
package com.naaturel.ANN.domain.model.neuron;
public class Input {
private float value;
public Input(float value){
this.value = value;
}
public void setValue(float value){
this.value = value;
}
public float getValue(){
return this.value;
}
}

30
src/main/java/com/naaturel/ANN/domain/model/neuron/Synapse.java Normal file
View File

@@ -0,0 +1,30 @@
package com.naaturel.ANN.domain.model.neuron;
public class Synapse {
private Input input;
private Weight weight;
public Synapse(Input input, Weight weight){
this.input = input;
this.weight = weight;
}
public float getInput(){
return this.input.getValue();
}
public void setInput(float value){
this.input.setValue(value);
}
public float getWeight() {
return weight.getValue();
}
public void setWeight(float value){
this.weight.setValue(value);
}
}

25
src/main/java/com/naaturel/ANN/domain/model/neuron/Weight.java Normal file
View File

@@ -0,0 +1,25 @@
package com.naaturel.ANN.domain.model.neuron;
import java.util.Random;
public class Weight {
private float value;
public Weight(){
this(new Random().nextFloat() * 2 - 1);
}
public Weight(float value){
this.value = value;
}
public void setValue(float value){
this.value = value;
}
public float getValue(){
return this.value;
}
}

17
src/main/java/com/naaturel/ANN/implementation/activationFunction/Heaviside.java Normal file
View File

@@ -0,0 +1,17 @@
package com.naaturel.ANN.implementation.activationFunction;
import com.naaturel.ANN.domain.abstraction.ActivationFunction;
import com.naaturel.ANN.domain.abstraction.Neuron;
public class Heaviside implements ActivationFunction {
public Heaviside(){
}
@Override
public float accept(Neuron n) {
float weightedSum = n.calculateWeightedSum();
return weightedSum <= 0 ? 0:1;
}
}

13
src/main/java/com/naaturel/ANN/implementation/activationFunction/Linear.java Normal file
View File

@@ -0,0 +1,13 @@
package com.naaturel.ANN.implementation.activationFunction;
import com.naaturel.ANN.domain.abstraction.ActivationFunction;
import com.naaturel.ANN.domain.abstraction.Neuron;
public class Linear implements ActivationFunction {
@Override
public float accept(Neuron n) {
return n.calculateWeightedSum();
}
}

32
src/main/java/com/naaturel/ANN/implementation/neuron/SimplePerceptron.java Normal file
View File

@@ -0,0 +1,32 @@
package com.naaturel.ANN.implementation.neuron;
import com.naaturel.ANN.domain.abstraction.ActivationFunction;
import com.naaturel.ANN.domain.abstraction.Neuron;
import com.naaturel.ANN.domain.abstraction.Trainable;
import com.naaturel.ANN.domain.model.neuron.Bias;
import com.naaturel.ANN.domain.model.neuron.Synapse;
import java.util.List;
public class SimplePerceptron extends Neuron implements Trainable {
public SimplePerceptron(List<Synapse> synapses, Bias b, ActivationFunction func) {
super(synapses, b, func);
}
@Override
public float predict() {
return activationFunction.accept(this);
}
@Override
public float calculateWeightedSum() {
float res = 0;
for(Synapse syn : super.synapses){
res += syn.getWeight() * syn.getInput();
}
return res;
}
}

4
src/main/java/com/naaturel/ANN/implementation/training/AdalineTraining.java Normal file
View File

@@ -0,0 +1,4 @@
package com.naaturel.ANN.implementation.training;
public class AdalineTraining {
}

99
src/main/java/com/naaturel/ANN/implementation/training/GradientDescentTraining.java Normal file
View File

@@ -0,0 +1,99 @@
package com.naaturel.ANN.implementation.training;
import com.naaturel.ANN.domain.abstraction.Neuron;
import com.naaturel.ANN.domain.model.dataset.DataSet;
import com.naaturel.ANN.domain.model.dataset.DataSetEntry;
import com.naaturel.ANN.domain.model.neuron.Bias;
import com.naaturel.ANN.domain.model.neuron.Input;
import com.naaturel.ANN.domain.model.neuron.Synapse;
import com.naaturel.ANN.domain.model.neuron.Weight;
import java.util.ArrayList;
import java.util.List;
public class GradientDescentTraining {
public GradientDescentTraining(){
}
public void train(Neuron n, float learningRate, DataSet dataSet) {
int epoch = 1;
int maxEpoch = 10000;
float errorThreshold = 0.125F;
float currentError;
do {
if(epoch > maxEpoch) break;
float biasCorrector = 0;
currentError = 0;
List<Float> correctorTerms = this.initCorrectorTerms(n.getSynCount());
for(DataSetEntry entry : dataSet) {
this.updateInputs(n, entry);
float prediction = n.predict();
float expectation = dataSet.getLabel(entry).getValue();
float delta = this.calculateDelta(expectation, prediction);
float loss = this.calculateLoss(delta);
currentError += loss;
Bias b = n.getBias();
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);
correctorTerms.set(i, c);
}
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 : %.2f]\n", currentError);
n.updateBias(new Weight(biasCorrector));
for(int i = 0; i < correctorTerms.size(); i++){
Synapse syn = n.getSynapse(i);
float c = correctorTerms.get(i);
syn.setWeight(syn.getWeight() + c);
}
epoch++;
} while(currentError > errorThreshold);
}
private List<Float> initCorrectorTerms(int number){
List<Float> res = new ArrayList<>();
for(int i = 0; i < number; i++){
res.add(0F);
}
return res;
}
private void updateInputs(Neuron n, DataSetEntry entry){
int index = 0;
for(float value : entry){
n.setInput(index, new Input(value));
index++;
}
}
private float calculateDelta(float expected, float predicted){
return expected - predicted;
}
private float calculateLoss(float delta){
return ((float) Math.pow(delta, 2))/2;
}
private float calculateWeightCorrection(float lr, float value, float delta){
return lr * value * delta;
}
}

68
src/main/java/com/naaturel/ANN/implementation/training/SimpleTraining.java Normal file
View File

@@ -0,0 +1,68 @@
package com.naaturel.ANN.implementation.training;
import com.naaturel.ANN.domain.abstraction.Neuron;
import com.naaturel.ANN.domain.model.dataset.DataSet;
import com.naaturel.ANN.domain.model.dataset.DataSetEntry;
import com.naaturel.ANN.domain.model.neuron.Input;
import com.naaturel.ANN.domain.model.neuron.Synapse;
import com.naaturel.ANN.domain.model.neuron.Weight;
public class SimpleTraining {
public SimpleTraining() {
}
public void train(Neuron n, float learningRate, DataSet dataSet) {
int epoch = 1;
int errorCount;
do {
errorCount = 0;
System.out.printf("Epoch : %d\n", epoch);
for(DataSetEntry entry : dataSet) {
this.updateInputs(n, entry);
float prediction = n.predict();
float expectation = dataSet.getLabel(entry).getValue();
float delta = this.calculateDelta(expectation, prediction);
float loss = this.calculateLoss(delta);
if(delta > 1e-6f) {
this.updateWeights(n, learningRate, delta);
errorCount += 1;
}
System.out.printf("predicted : %.2f, ", prediction);
System.out.printf("expected : %.2f, ", expectation);
System.out.printf("delta : %.2f\n", this.calculateDelta(expectation, prediction));
}
System.out.print("====================================\n");
epoch++;
} while (errorCount != 0);
}
private void updateInputs(Neuron n, DataSetEntry entry){
int index = 0;
for(float value : entry){
n.setInput(index, new Input(value));
index++;
}
}
private void updateWeights(Neuron n, float rate, float delta){
Weight biasCorrection = new Weight(n.getBias().getWeight() + (rate * delta * n.getBias().getInput()));
n.updateBias(biasCorrection);
for(Synapse syn : n.getSynapses()){
syn.setWeight(syn.getWeight() + (rate * delta * syn.getInput()));
}
}
private float calculateDelta(float expected, float predicted){
return expected - predicted;
}
private float calculateLoss(float delta){
return Math.abs(delta);
}
}