Contents
Preparation
Follow the instructions from the assignment Creative Programming to prepare the software environment, including Processing and Arduino. If you are not planning to use the GUI components from the ControlP5 library, you can skip the ControlP5 part.
Install Neuroph, a lightweight Java neural network framework.
- Download proper neurophstudio installer for your platform. I will use Windows platform as examples here
- In "Processing Sketchbook location"\libraries, create a sub-directory "neuroph".
- In "Processing Sketchbook location"\libraries\neuroph, create a sub-directory "library".
From <neurophstudio directory>\neurophstudio\modules\ext copy neuroph-core-<version>.jar to <Processing Sketchbook location>\libraries\neuroph\library. Rename neuroph-core-<version>.jar to neuroph.jar.
If you are confident enough, you can also try to use Eclipse to program in Java. Then try to follow the instructions from the assignment Processing2Java. Remember you shall include the aforementioned jar files in the projects if you want to use neuroph.
References
Examples
Try the Perceptron example first with Neuroph Studio: This is the Perceptron example from http://neuroph.sourceforge.net/tutorials/Perceptron.html:
Java code
1 package org.neuroph.samples;
2
3 import org.neuroph.core.NeuralNetwork;
4 import org.neuroph.nnet.Perceptron;
5 import org.neuroph.core.learning.DataSet;
6 import org.neuroph.core.learning.TrainingElement;
7 import org.neuroph.core.learning.SupervisedTrainingElement;
8 import java.util.Vector;
9
10 /**
11 * This sample shows how to create, train, save and load simple Perceptron neural network
12 */
13 public class PerceptronSample {
14
15 public static void main(String args[]) {
16
17 // create training set (logical AND function)
18 DataSet trainingSet = new DataSet(2, 1);
19 trainingSet.addRow(new DataSetRow(new double[] {
20 0, 0
21 }
22 , new double[] {
23 0
24 }
25 ));
26 trainingSet.addRow(new DataSetRow(new double[] {
27 0, 1
28 }
29 , new double[] {
30 0
31 }
32 ));
33 trainingSet.addRow(new DataSetRow(new double[] {
34 1, 0
35 }
36 , new double[] {
37 0
38 }
39 ));
40 trainingSet.addRow(new DataSetRow(new double[] {
41 1, 1
42 }
43 , new double[] {
44 1
45 }
46 ));
47
48 // create perceptron neural network
49 NeuralNetwork myPerceptron = new Perceptron(2, 1);
50
51 // learn the training set
52 myPerceptron.learn(trainingSet);
53
54 // test perceptron
55 System.out.println("Testing trained perceptron");
56 testNeuralNetwork(myPerceptron, trainingSet);
57
58 // save trained perceptron
59 myPerceptron.save("mySamplePerceptron.nnet");
60
61 // load saved neural network
62 NeuralNetwork loadedPerceptron = NeuralNetwork.load("mySamplePerceptron.nnet");
63
64 // test loaded neural network
65 System.out.println("Testing loaded perceptron");
66 testNeuralNetwork(loadedPerceptron, trainingSet);
67 }
68
69 public static void testNeuralNetwork(NeuralNetwork nnet, DataSet tset) {
70
71 for (DataSetRow dataRow : tset.getRows()) {
72
73 nnet.setInput(dataRow.getInput());
74 nnet.calculate();
75 double[ ] networkOutput = nnet.getOutput();
76 System.out.print("Input: " + Arrays.toString(dataRow.getInput()) );
77 System.out.println(" Output: " + Arrays.toString(networkOutput) );
78 }
79 }
80 }
Processing code
1 import org.neuroph.util.plugins.*;
2 import org.neuroph.core.transfer.*;
3 import org.neuroph.util.*;
4 import org.neuroph.nnet.comp.layer.*;
5 import org.neuroph.nnet.*;
6 import org.neuroph.core.events.*;
7 import org.neuroph.core.learning.error.*;
8 import org.neuroph.core.data.sample.*;
9 import org.neuroph.util.io.*;
10 import org.neuroph.core.exceptions.*;
11 import org.neuroph.core.learning.*;
12 import org.neuroph.core.input.*;
13 import org.neuroph.nnet.learning.*;
14 import org.neuroph.util.benchmark.*;
15 import org.neuroph.util.random.*;
16 import org.neuroph.nnet.comp.*;
17 import org.neuroph.core.learning.stop.*;
18 import org.neuroph.core.data.*;
19 import org.neuroph.core.*;
20 import org.neuroph.nnet.comp.neuron.*;
21 import org.neuroph.core.data.norm.*;
22
23 import java.util.*;
24
25 /**
26 * This sample shows how to create, train, save and load simple Perceptron neural network
27 */
28 void setup() {
29
30 // create training set (logical AND function)
31 DataSet trainingSet = new DataSet(2, 1);
32 trainingSet.addRow(new DataSetRow(new double[] {
33 0, 0
34 }
35 , new double[] {
36 0
37 }
38 ));
39 trainingSet.addRow(new DataSetRow(new double[] {
40 0, 1
41 }
42 , new double[] {
43 0
44 }
45 ));
46 trainingSet.addRow(new DataSetRow(new double[] {
47 1, 0
48 }
49 , new double[] {
50 0
51 }
52 ));
53 trainingSet.addRow(new DataSetRow(new double[] {
54 1, 1
55 }
56 , new double[] {
57 1
58 }
59 ));
60
61 // create perceptron neural network
62 NeuralNetwork myPerceptron = new Perceptron(2, 1);
63
64 // learn the training set
65 myPerceptron.learn(trainingSet);
66
67 // test perceptron
68 System.out.println("Testing trained perceptron");
69 testNeuralNetwork(myPerceptron, trainingSet);
70
71 // save trained perceptron
72 myPerceptron.save("mySamplePerceptron.nnet");
73
74 // load saved neural network
75 NeuralNetwork loadedPerceptron = NeuralNetwork.load("mySamplePerceptron.nnet");
76
77 // test loaded neural network
78 System.out.println("Testing loaded perceptron");
79 testNeuralNetwork(loadedPerceptron, trainingSet);
80 }
81
82 void testNeuralNetwork(NeuralNetwork nnet, DataSet tset) {
83
84 for (DataSetRow dataRow : tset.getRows()) {
85
86 nnet.setInput(dataRow.getInput());
87 nnet.calculate();
88 double[ ] networkOutput = nnet.getOutput();
89 System.out.print("Input: " + Arrays.toString(dataRow.getInput()) );
90 System.out.println(" Output: " + Arrays.toString(networkOutput) );
91 }
92 }