[marf-cvs] marf/src/marf/Classification/NeuralNetwork Neuron.java,NONE,1.1

[<icl...@pr...>]

Update of /cvsroot/marf/marf/src/marf/Classification/NeuralNetwork
In directory sc8-pr-cvs1:/tmp/cvs-serv17380

Added Files:
	Neuron.java 
Log Message:
DEV: From /resources/NeuralNetwork/


--- NEW FILE: Neuron.java ---
import java.util.ArrayList;
import java.lang.Double;
import java.lang.Math;

public class Neuron {

    //Data Members
    public String name;
    public int type = -1; //input=0; hidden=1; output=2; undef=-1;

    private ArrayList inputs = new ArrayList();

    private ArrayList weights = new ArrayList();
    private ArrayList weightsBuf = new ArrayList();
    
    private ArrayList outputs = new ArrayList();

    public double delta = 0.0;
    public double threshold = 0.0;
    public double result = 0.0;

    Neuron(String n, int t) {
	name = new String(n);
	type = t;
    }

    //Methods

    public boolean addInput(Neuron in, double weight) {
	return inputs.add(in) && 
	    weights.add(new Double(weight)) && 
	    weightsBuf.add(new Double(weight));
    }

    public boolean addOutput(Neuron out) {
	return outputs.add(out);
    }

    public void eval() {

	if(type == 0) return; //assumes that the result of an input neuron is == the input

	if(inputs.isEmpty()) return;
	
	double count = 0;

	for(int i=0; i<inputs.size(); i++)
	    count += ((Neuron)inputs.get(i)).result * ((Double)weights.get(i)).doubleValue();

	count -= threshold;

	result = 1.0 / (1.0 + Math.exp(-count));

	//System.out.println("Neuron: " + name + ", Sum: " + count + ", Result: " + result);
    }
    
    private double getWeight(Neuron n) {
	int val = inputs.indexOf(n);
	
	if(val>=0) 
	    return ((Double)weights.get(val)).doubleValue();
	
	//System.out.println("There is no pointer n in neuron");
	
	return -1.0;
    }

    public void train(double expected, double alpha, double beta) {

	if(type == 2) { //output nodes calc delta differntly based on expected res...
	    delta = (expected - result) * result * (1.0 - result);
	}
	else if(type == 1) {
	    
	    double sum = 0.0;
	
	    for(int i=0; i<outputs.size(); i++)
		sum += ((Neuron)outputs.get(i)).delta * ((Neuron)outputs.get(i)).getWeight(this);
	
	    delta = result * (1.0 - result) * sum;
	}
	else {
	    return;
	}

	//System.out.println("Neuron: " + name + ", Delta: " + delta);

	for(int i=0; i<inputs.size(); i++) {
	    weightsBuf.set(i, new Double(beta * ((Double)weights.get(i)).doubleValue() + 
					 alpha * delta * ((Neuron)inputs.get(i)).result));

	    //double arg = ((Double)weights.get(i)).doubleValue();
	    //System.out.println("\tNew weight " + i + ": " + arg);
	}

    }

    public void commit() {
	for(int i=0; i<weights.size(); i++) {
	    weights.set(i, new Double(((Double)weightsBuf.get(i)).doubleValue()));
	}
    }

    public void printXML(int tab) {
	
	indent(tab);
	System.out.println("<neuron index=\"" + name + "\" thresh=\"" + threshold + "\">");
        
	for(int i=0; i<inputs.size(); i++) {
	    indent(tab+1);
	    System.out.println("<input ref=\"" + ((Neuron)inputs.get(i)).name + "\" weight=\"" + (Double)weights.get(i) + "\"/>");
	}

	for(int i=0; i<outputs.size(); i++) {
	    indent(tab+1);
	    System.out.println("<output ref=\"" + ((Neuron)outputs.get(i)).name + "\"/>");
	}

	indent(tab);
	System.out.println("</neuron>");
    }

    public void indent(int n) {
	for(int i=0; i<n; i++) System.out.print("\t");
    }

}