airwin.java

All the tool for build a network able to reconize any shapes. Very complete, a good base for anythi

/*
Coded by Aydin Gurel, 2003
The code is free, but please contact me if you wish to use the code entirely or partially in any kind of project so that I can reference it and please don't delete these lines so that other people can reach this information. Also, please inform me if you encounter a bug.
[e-mail]
http://aydingurel.brinkster.com/neural

The purpose of this example is to show you how this package should be used, rather than to give you clues about how to train and use a neural network in general.

In this example, we:
-aim to resolve a classification problem,
-create a multilayer perceptron,
-save the configuration of the net,
-create three patternsets;
-one for training, one for cross validation and one for testing.
-train the net using mini batch, until crossvalerror < 0.02, so that it learns how to distinguish a female crab by using some measurements,
-save its weights,
-recreate the net using previously saved configuration and weigts,
-test it

training data was taken from a tutorial of the "NeuroSolutions" software
http://www.neurosolutions.com/
 */

import java.io.IOException;

import jnt.FFT.ComplexDouble2DFFT;

import neuralNetwork.*;
import mouse.*;

public class AirWin {
	public static void main(String args[]) {

		mouseTracker myMouseTracker = new mouseTracker();
		Randomizer randomizer = new Randomizer();
		int size=34;
		
		// create a multilayer perceptron with four layers:
		// one input layer with seven units; two hidden layers
		// each with ten units, using tanh function;
		// one output layer with one unit using tanh function.
		// except for noofneurons, all parameters for the input layer
		// are ineffectual.
		int[] noofneurons = { size * size * 2, 3000, 4 };
		double[] learnratecoeff = { .8, .8, .8 };
		char[] axonfamily = { 't', 't', 't' };
		double[] momentumrate = { 0, 1.2, 1.1 };
		double[] flatness = { 1, 1.2, 1 };
		System.out.println("Creating the net");
		NeuralNet mynet = new NeuralNet(noofneurons, learnratecoeff,
				axonfamily, momentumrate, flatness, randomizer);
		// Save the configuration to a file
		System.out.println("Saving the configuration");
		try {
			mynet.SaveConfig("AirWin.nnc");
		} catch (IOException e) {
		}

		// create three pattern sets with 7 input and 1 output values.
		System.out.println("Loading patterns");
		// first create a pattern set for training
		PatternSet trainingpatterns = new PatternSet("AirWin_training.csv",
				 size * size * 2, 4, 1,
				0, 0, randomizer);
		// then create a pattern set for cross validation
		PatternSet crossvalpatterns = new PatternSet("AirWin_crossval.csv",
				 size * size * 2, 4, 0,
				1, 0, randomizer);
		// and then create a pattern set for testing
		PatternSet testpatterns = new PatternSet("AirWin_test.csv",
				 size * size * 2, 4, 0,
				0, 1, randomizer);

		// show the error ratio before training
		System.out.println("Error ratio before training: "
				+ mynet.CrossValErrorRatio(crossvalpatterns));

		// train the net using mini batch training
		System.out.println("Beginning mini batch training");
		double temp_err;
		temp_err = mynet.CrossValErrorRatio(crossvalpatterns);
		while (temp_err > 0.3) {
			System.out.println("Training the net. Error ratio: " + temp_err);
			mynet.MinibatchTrainPatterns(trainingpatterns.trainingpatterns, .3,
					30);
			temp_err = mynet.CrossValErrorRatio(crossvalpatterns);
		}

		/*
		 * // or you can use incremental training:
		 * System.out.println("Beginning incremental training"); double
		 * temp_err; temp_err = mynet.CrossValErrorRatio(crossvalpatterns);
		 * while (temp_err > .3) {
		 * System.out.println("Training the net. Error ratio: " + temp_err );
		 * mynet.IncrementalTrainPatterns(trainingpatterns.trainingpatterns,
		 * .01); temp_err = mynet.CrossValErrorRatio(crossvalpatterns); }
		 */

		// finally, check the error using test data
		System.out.println("Error ratio of the test data: "
				+ mynet.TestErrorRatio(testpatterns));

		System.out.println("Training is over");

		// now that the training is over, save the weights of the net.
		System.out.println("Saving the weights\n");
		try {
			mynet.SaveWeights("AirWin.nnw");
		} catch (IOException e) {
		}

		// clean up the objects
		trainingpatterns = null;
		crossvalpatterns = null;
		testpatterns = null;
		mynet = null;
		randomizer = null;

		// /痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/痋_/

		// now recreate the net using previously saved data and
		// test it.

		// recreate the net
		randomizer = new Randomizer();
		System.out.println("Recreating the net");
		mynet = new NeuralNet("AirWin.nnc", randomizer);
		mynet.LoadWeights("AirWin.nnw");
		
		int ind;
		
		double[] newPict= new double[myMouseTracker.yResolution*myMouseTracker.xResolution*2];
		double[] transPict= new double[size*size*2];
		ComplexDouble2DFFT methodes = new ComplexDouble2DFFT(myMouseTracker.xResolution, myMouseTracker.yResolution);
		
		// and test it
		while (true) {
			myMouseTracker.fill();
			ind=0;
			for (int j=0; j<myMouseTracker.yResolution; j++){
				for (int k=0; k<myMouseTracker.xResolution; k++){
					newPict[ind++]=myMouseTracker.binaryMap[k][j];
					newPict[ind++]=0;
				}
			}
		
		methodes.transform(newPict);
		
		ind = 0;
		for (int k=0; k<newPict.length ;k++){
			if ((k%myMouseTracker.xResolution*2<=size || k%myMouseTracker.xResolution*2>myMouseTracker.xResolution*2-size) && (k<(myMouseTracker.xResolution*2)*(size/2) || k>=(myMouseTracker.xResolution*myMouseTracker.yResolution*2)-(myMouseTracker.xResolution*2)*(size/2))){
				transPict[ind++]=newPict[k];
			}
		}
			double[] outputs = mynet.Output(transPict);
			if (outputs[0] > 0)
				System.out.println("Result: Rond");

			if (outputs[1] > 0)
				System.out.println("Result: croix");

			if (outputs[2] > 0)
				System.out.println("Result: barres horizontales");

			if (outputs[3] > 0)
				System.out.println("Result: barres verticales");
		}
		// mynet = null;
	}
}