import java.applet.Applet; import java.awt.*; import java.awt.event.*; import java.util.*; import java.io.*; public class MLP_Applet extends Applet implements Runnable, ActionListener { //private members //Threads Thread myThread; //panels private Panel ap0, ap1, ap2, ap3, ap4; private Panel p1, p2, p3, p4, p5, p6, p7, p8, p9, p10; private Panel p11, p12, p13, p14, p15, p16, p17; //frames private Frame mainFrame, paramFrame, featureFrame, targetFrame, valFrame; private Frame gFrame1, gFrame2, helpFrame; //labels private Label title1, title2, title3, title4; private Label l1, l2, l3, l4, l5, l6, l7; //buttons private Button MLP_Param; private Button testVec, trainVec, valVec; private Button Default_Param, submit, submit1, submit2, submit3; private Button train, test, validate, clear, errorStat; private Button param, graph1, graph2; private Button close1, close2, stopBtn, close3, clear1, clear2, clear3; //text fields private TextField tf1, tf2, tf3, tf4, tf5, tf6, tf7, tf8, tf9, tf10; private TextField tf11, tf12, tf13, tf14, tf15, tf16; //text area private TextArea featureArea, featureArea1, targetArea, targetArea1, errorArea; private TextArea valFeatureArea, valTargetArea, helpArea; //radio buttons private CheckboxGroup sigmoid, range; private Checkbox unipolar, bipolar, rmin, rmax; //Menu bars private MenuBar main, fmbar1, fmbar2, fmbar3, fmbar4; private Menu fileMenu, helpMenu, fm1, fm2, fm3, fm4, hm1, hm2, hm3, hm4; private MenuItem i1, i2, i3, i4, i4a, i5, i6, i7, i7a, i8, i9, i10, i11; private MenuItem i12, i13, i14, i15; //File dialogues private FileDialog fileLoader, fileSaver; //variables - strings private String N, M, J, Q, valQ, testQ, a1, a2, da1, da2, b1, b2, e1, e2, mu, num; private String featureVectors, featureVectors1, targetVectors, targetVectors1; private String valFeatureVectors, valTargetVectors; private String iteStr, errStr; //variables string tokenizers private StringTokenizer fv, fv1, tv, tv1, val1, val2; //variables boolean private boolean frozen, valFlag, grph1Flag, grph2Flag; //variables - integers private int vN, vM, vJ, vQ, vtestQ, vvalQ, vnum; private int iteration; //variables - double private double va1, va2, vda1, vda2, vah, vb1, vb2, ve1, ve2, vmu; private double dgds, dhdr, sumW; private double oldErr, newErr, ssErr, oldValErr, newValErr, verErr; //variables - double arrays private double X[][], testX[][], valX[][], T[][], testT[][], valT[][]; private double W[][], U[][], Y[][], testY[][], valY[][], Z[][], testZ[][], valZ[][]; private double delW[][], ndelW[][], delU[][], ndelU[][]; private double rSum[][], testrSum[][], valrSum[][], sSum[][], testsSum[][], valsSum[][]; private double Err[], temp[], Zgrph[], Tgrph[]; //graph instances Graph1 G1; Graph2 G2; Graphics grph1, grph2; //Graph vaiables String g1str1, g2str1; int g1px1, g2px1, g1px2, g2px2, g1py1, g2py1, g1py2, g2py2; int g1lastx, g1lasty, g2lastx, g2lasty; double g1ymax, g1ymin, g2ymax, g2ymin; //Get min and max x-axis( iterations )/y-axis ( error ) double g1xmin, g2xmin; double g1xmax, g2xmax; double g2ymin1; double g2ymax1; double g2ymin2; double g2ymax2; //determine x-axis tick marks double g1xtic1, g1xtic2, g1xtic3, g1xtic4; double g2xtic1, g2xtic2, g2xtic3, g2xtic4; //determine y-axis tick marks double g1ytic1, g1ytic2, g1ytic3, g1ytic4; double g2ytic1, g2ytic2, g2ytic3, g2ytic4; public void init() { valFlag = true; grph1Flag = false; grph2Flag = false; //Create a parent frame to hold applet widgets mainFrame = new Frame( "Welcome to my MLPs Applet" ); mainFrame.setLayout( new FlowLayout() ); //MENU stuff... for main frame main = new MenuBar(); fileMenu = new Menu( "File" ); helpMenu = new Menu( "Help" ); i1 = new MenuItem( "Save" ); fileMenu.add( i1 ); i1.addActionListener( this ); i2 = new MenuItem( "Quit" ); fileMenu.add( i2 ); i2.addActionListener( this ); main.add( fileMenu ); helpMenu = new Menu( "Help" ); i3 = new MenuItem( "Using this applet" ); helpMenu.add( i3 ); i3.addActionListener( this ); main.add( helpMenu ); mainFrame.setMenuBar( main ); //Add a button for input parameters ap0 = new Panel(); ap0.setLayout( new FlowLayout() ); MLP_Param = new Button( "Input MLPs Parameters" ); MLP_Param.addActionListener( this ); ap0.add( MLP_Param ); mainFrame.add( ap0 ); //Add buttons for input vectors for training, testing, and validation ap1 = new Panel(); ap1.setLayout( new FlowLayout() ); trainVec = new Button( "Input Vectors for Training" ); trainVec.addActionListener( this ); ap1.add( trainVec ); testVec = new Button( "Input Vectors for Testing" ); testVec.addActionListener( this ); ap1.add( testVec ); valVec = new Button( "Input Vectors for Validation" ); valVec.addActionListener( this ); ap1.add( valVec ); mainFrame.add( ap1 ); //Add error text area in this applet panel ap2 = new Panel(); ap2.setLayout( new FlowLayout() ); errorArea = new TextArea(); ap2.add( errorArea ); mainFrame.add( ap2 ); //Add training/testing/view buttons in this applet panel ap3 = new Panel(); ap3.setLayout( new FlowLayout() ); train = new Button( "Start Training" ); train.addActionListener( this ); ap3.add( train ); stopBtn = new Button( "Stop" ); stopBtn.addActionListener( this ); ap3.add( stopBtn ); validate = new Button( "Validate NN" ); validate.addActionListener( this ); ap3.add( validate ); test = new Button( "Start Testing" ); test.addActionListener( this ); ap3.add( test ); errorStat = new Button( "View Error Statistics" ); errorStat.addActionListener( this ); ap3.add( errorStat ); clear = new Button( "Refresh" ); clear.addActionListener( this ); ap3.add( clear ); mainFrame.add( ap3 ); //Add graph buttons here ap4 = new Panel(); ap4.setLayout( new FlowLayout() ); param = new Button( "View Learned Parameters" ); param.addActionListener( this ); ap4.add( param ); graph1 = new Button( "View Error Graph" ); graph1.addActionListener( this ); ap4.add( graph1 ); graph2 = new Button( "View Prediction Graph" ); graph2.addActionListener( this ); ap4.add( graph2 ); mainFrame.add( ap4 ); mainFrame.setSize( 530, 425 ); mainFrame.setVisible( true ); //Create a frame to hold input parameters paramFrame = new Frame( "MLPs Parameter" ); paramFrame.setLayout( new GridLayout( 14, 1 ) ); helpFrame = new Frame( "MLPs Parameter Help" ); helpFrame.setLayout( new FlowLayout() ); //Create a frame to hold feature and target vectors for training featureFrame = new Frame( "Input Vectors for Training" ); featureFrame.setLayout( new GridLayout( 1, 1 ) ); //Create a frame to hold feature and target vectors for testing targetFrame = new Frame( "Input Vectors for Testing" ); targetFrame.setLayout( new GridLayout( 1, 1 ) ); //Create a frame to hold feature and target vectors for testing valFrame = new Frame( "Input Vectors for Validation" ); valFrame.setLayout( new GridLayout( 1, 1 ) ); //Panel for N p1 = new Panel(); p1.setLayout( new FlowLayout() ); p1.add( new Label( "Number of Input Nodes ( N ) " ) ); tf1 = new TextField( 3 ); tf1.addActionListener( this ); p1.add( tf1 ); paramFrame.add( p1 ); //Panel for M p2 = new Panel(); p2.setLayout( new FlowLayout() ); p2.add( new Label( "Number of Middle Neurodes ( M )" ) ); tf2 = new TextField( 3 ); tf2.addActionListener( this ); p2.add( tf2 ); paramFrame.add( p2 ); //Panel for J p3 = new Panel(); p3.setLayout( new FlowLayout() ); p3.add( new Label( "Number of Output Neurodes ( J ) " ) ); tf3 = new TextField( 3 ); tf3.addActionListener( this ); p3.add( tf3 ); paramFrame.add( p3 ); //Panel for Q p4 = new Panel(); p4.setLayout( new FlowLayout() ); p4.add( new Label( "Number of Exemplar Vectors ( Q )" ) ); tf4 = new TextField( 3 ); tf4.addActionListener( this ); p4.add( tf4 ); paramFrame.add( p4 ); //Panel for sigmoid function label p5 = new Panel(); p5.setLayout( new FlowLayout() ); title1 = new Label( "Select one of the sigmoid functions" ); title1.setFont( new Font( "TimesRoman", Font.BOLD, 14 ) ); p5.add( title1 ); paramFrame.add( "Center", p5 ); //Panel for sigmoid function radio buttons p6 = new Panel(); p6.setLayout( new FlowLayout() ); sigmoid = new CheckboxGroup(); unipolar = new Checkbox( "Unipolar Sigmoid", sigmoid, false ); bipolar = new Checkbox( "Bipolar Sigmoid", sigmoid, false ); p6.add( unipolar ); p6.add( bipolar ); paramFrame.add( "Center", p6 ); //Panel for sigmoid parameters label p7 = new Panel(); p7.setLayout( new FlowLayout() ); title2 = new Label( "Sigmoid Parameters" ); title2.setFont( new Font( "TimesRoman", Font.BOLD, 14 ) ); p7.add( title2 ); paramFrame.add( "Center", p7 ); //Panel for sigmoid parameters alpha1, alpha2, and deltas p8 = new Panel(); p8.setLayout( new FlowLayout() ); p8.add( new Label( "alpha1" ) ); tf5 = new TextField( 3 ); tf5.addActionListener( this ); p8.add( tf5 ); p8.add( new Label( "delta_alpha1" ) ); tf6 = new TextField( 3 ); tf6.addActionListener( this ); p8.add( tf6 ); p8.add( new Label( "bias1 = N/" ) ); tf7 = new TextField( 3 ); tf7.addActionListener( this ); p8.add( tf7 ); paramFrame.add( "Center", p8 ); //Panel for sigmoid parameters bias1 and bias2 p9 = new Panel(); p9.setLayout( new FlowLayout() ); p9.add( new Label( "alpha2" ) ); tf8 = new TextField( 3 ); tf8.addActionListener( this ); p9.add( tf8 ); p9.add( new Label( "delta_alpha2" ) ); tf9 = new TextField( 3 ); tf9.addActionListener( this ); p9.add( tf9 ); p9.add( new Label( "bias2 = M/" ) ); tf10 = new TextField( 3 ); tf10.addActionListener( this ); p9.add( tf10 ); paramFrame.add( "Center", p9 ); //Panel for step gain and momentum label p10 = new Panel(); p10.setLayout( new FlowLayout() ); title3 = new Label( "Step gain and momentum parameters" ); title3.setFont( new Font( "TimesRoman", Font.BOLD, 14 ) ); p10.add( title3 ); paramFrame.add( "Center", p10 ); //Panel for step gain and momentum parameters p11 = new Panel(); p11.setLayout( new FlowLayout() ); p11.add( new Label( "eta1" ) ); tf11 = new TextField( 3 ); tf11.addActionListener( this ); p11.add( tf11 ); p11.add( new Label( "eta2" ) ); tf12 = new TextField( 3 ); tf12.addActionListener( this ); p11.add( tf12 ); p11.add( new Label( "momentum" ) ); tf13 = new TextField( 3 ); tf13.addActionListener( this ); p11.add( tf13 ); paramFrame.add( "Center", p11 ); //Panel for number of iterations p12 = new Panel(); p12.setLayout( new FlowLayout() ); title4 = new Label( "Number of Iterations" ); title4.setFont( new Font( "TimesRoman", Font.BOLD, 14 ) ); p12.add( title4 ); tf14 = new TextField( 3 ); tf14.addActionListener( this ); p12.add( tf14 ); paramFrame.add( "Center", p12 ); //panel for default and previous parameter buttons p13 = new Panel(); p13.setLayout( new FlowLayout() ); Default_Param = new Button( "View Default Parameters" ); Default_Param.addActionListener( this ); p13.add( Default_Param ); paramFrame.add( "Center", p13 ); //panel for submit parameter button p14 = new Panel(); p14.setLayout( new FlowLayout() ); submit = new Button( "Submit" ); submit.addActionListener( this ); p14.add( submit ); paramFrame.add( "Center", p14 ); fmbar3 = new MenuBar(); fm3 = new Menu( "File" ); i10 = new MenuItem( "Close" ); fm3.add( i10 ); i10.addActionListener( this ); fmbar3.add( fm3 ); hm3 = new Menu( "Help" ); i11 = new MenuItem( "Using this window" ); hm3.add( i11 ); i11.addActionListener( this ); fmbar3.add( hm3 ); paramFrame.setMenuBar( fmbar3 ); //panel for feature/target vectors for training and submit1 button p15 = new Panel(); p15.setLayout( new FlowLayout() ); fmbar1 = new MenuBar(); fm1 = new Menu( "File" ); i4 = new MenuItem( "Feature Vectors" ); fm1.add( i4 ); i4.addActionListener( this ); i4a = new MenuItem( "Target Vectors" ); fm1.add( i4a ); i4a.addActionListener( this ); i5 = new MenuItem( "Close" ); fm1.add( i5 ); i5.addActionListener( this ); fmbar1.add( fm1 ); hm1 = new Menu( "Help" ); i6 = new MenuItem( "Using this window" ); hm1.add( i6 ); i6.addActionListener( this ); fmbar1.add( hm1 ); featureFrame.setMenuBar( fmbar1 ); // range = new CheckboxGroup(); // rmin = new Checkbox( "Map data from 0.1 to 0.9", range, false ); // rmax = new Checkbox( "Map data from -0.9 to 0.9", range, false ); // p15.add( rmin ); // p15.add( rmax ); l1 = new Label( "Enter Input Feature and Target Vectors" ); p15.add( l1 ); featureArea = new TextArea(); p15.add( featureArea ); //l2 = new Label( "Enter Input Target Vectors" ); //p15.add( l2 ); featureArea1 = new TextArea(); p15.add( featureArea1 ); submit1 = new Button( "Submit Vectors for Training" ); submit1.addActionListener( this ); p15.add( submit1 ); clear1 = new Button( "Clear" ); clear1.addActionListener( this ); p15.add( clear1 ); featureFrame.add( "Center", p15 ); //panel for target/feature vectors for testing and submit2 button p16 = new Panel(); p16.setLayout( new FlowLayout() ); fmbar2 = new MenuBar(); fm2 = new Menu( "File" ); i7 = new MenuItem( "Feature Vectors" ); fm2.add( i7 ); i7.addActionListener( this ); i7a = new MenuItem( "Target Vectors" ); fm2.add( i7a ); i7a.addActionListener( this ); i8 = new MenuItem( "Close" ); fm2.add( i8 ); i8.addActionListener( this ); fmbar2.add( fm2 ); hm2 = new Menu( "Help" ); i9 = new MenuItem( "Using this window" ); hm2.add( i9 ); i9.addActionListener( this ); fmbar2.add( hm2 ); targetFrame.setMenuBar( fmbar2 ); p16.add( new Label( "Enter Number of Exemplar Vectors ( Q ) " ) ); tf15 = new TextField( 3 ); tf15.addActionListener( this ); p16.add( tf15 ); l3 = new Label( "Enter Input Feature and Target Vectors" ); p16.add( l3 ); targetArea = new TextArea(); p16.add( targetArea ); //l4 = new Label( "Enter Input Target Vectors" ); //p16.add( l4 ); targetArea1 = new TextArea(); p16.add( targetArea1 ); submit2 = new Button( "Submit Vectors for Testing" ); submit2.addActionListener( this ); p16.add( submit2 ); clear2 = new Button( "Clear" ); clear2.addActionListener( this ); p16.add( clear2 ); targetFrame.add( "Center", p16 ); //panel for target/feature vectors for validation and submit3 button p17 = new Panel(); p17.setLayout( new FlowLayout() ); fmbar4 = new MenuBar(); fm4 = new Menu( "File" ); i12 = new MenuItem( "Feature Vectors" ); fm4.add( i12 ); i12.addActionListener( this ); i13 = new MenuItem( "Target Vectors" ); fm4.add( i13 ); i13.addActionListener( this ); i14 = new MenuItem( "Close" ); fm4.add( i14 ); i14.addActionListener( this ); fmbar4.add( fm4 ); hm4 = new Menu( "Help" ); i15 = new MenuItem( "Using this window" ); hm4.add( i15 ); i15.addActionListener( this ); fmbar4.add( hm4 ); valFrame.setMenuBar( fmbar4 ); p17.add( new Label( "Enter Number of Exemplar Vectors ( Q ) " ) ); tf16 = new TextField( 3 ); tf16.addActionListener( this ); p17.add( tf16 ); l4 = new Label( "Enter Input Feature and Target Vectors" ); p17.add( l4 ); valFeatureArea = new TextArea(); p17.add( valFeatureArea ); valTargetArea = new TextArea(); p17.add( valTargetArea ); submit3 = new Button( "Submit Vectors for Validation" ); submit3.addActionListener( this ); p17.add( submit3 ); clear3 = new Button( "Clear" ); clear3.addActionListener( this ); p17.add( clear3 ); valFrame.add( "Center", p17 ); //paramFrame help section helpArea = new TextArea(); helpFrame.add( helpArea ); close3 = new Button( "Close" ); close3.addActionListener( this ); helpFrame.add( close3 ); //Graphs gFrame1 = new Frame( "Error Graph" ); gFrame1.setLayout( new FlowLayout() ); G1 = new Graph1( Color.lightGray, 350, 350 ); close1 = new Button( "Close" ); close1.addActionListener( this ); gFrame1.add( G1 ); gFrame1.add( close1 ); gFrame2 = new Frame( "Prediction Graph" ); gFrame2.setLayout( new FlowLayout() ); G2 = new Graph2( Color.lightGray, 350, 350 ); close2 = new Button( "Close" ); close2.addActionListener( this ); gFrame2.add( G2 ); gFrame2.add( close2 ); } //Event Handling public void actionPerformed( ActionEvent event ) { if( event.getSource() == MLP_Param ){ paramFrame.setSize( 475, 475 ); paramFrame.setVisible( true ); } else if( event.getSource() == i10 ){ paramFrame.setVisible( false ); } else if( event.getSource() == i11 ){ //Display help on using paramFrame helpFrame.setSize( 475, 275 ); helpFrame.setVisible( true ); paramFrameHelp(); } else if( event.getSource() == close3 ){ helpFrame.setVisible( false ); } else if( event.getSource() == clear1 ){ featureArea.setText( "" ); featureArea1.setText( "" ); } else if( event.getSource() == clear2 ){ targetArea.setText( "" ); targetArea1.setText( "" ); } else if( event.getSource() == clear3 ){ valFeatureArea.setText( "" ); valTargetArea.setText( "" ); } else if( event.getSource() == i1 ){ try{ fileSaver = new FileDialog( mainFrame, "Browse", FileDialog.SAVE ); fileSaver.show(); saveFile( fileSaver.getFile() ); } catch( SecurityException se3 ){ errorArea.setText( "Sorry Could open file dialog box. You may want to copy\nthe text area content and paste it in a text file.\n" + se3 ); } } else if( event.getSource() == i2 ) { mainFrame.setVisible( false ); } else if( event.getSource() == i3 ){ //Display help using this applet MainHelp(); } else if( event.getSource() == i4 ){ try{ fileLoader = new FileDialog( mainFrame, "Browse", FileDialog.LOAD ); fileLoader.show(); loadFile( fileLoader.getFile(), "i4" ); } catch( SecurityException se4 ){ errorArea.setText( "Sorry Could open file dialog box. You may want to copy the\nfeature vectors for training and paste it in the text area.\n" + se4 ); } } else if( event.getSource() == i4a ){ try{ fileLoader = new FileDialog( mainFrame, "Browse", FileDialog.LOAD ); fileLoader.show(); loadFile( fileLoader.getFile(), "i4a" ); } catch( SecurityException se5 ){ errorArea.setText( "Sorry Could open file dialog box. You may want to copy the\ntarget vectors for training and paste it in the text area.\n" + se5 ); } } else if( event.getSource() == i5 ){ featureFrame.setVisible( false ); } else if( event.getSource() == i6 ){ //Display help using featureFrame ( training vectors ) trainVecHelp(); } else if( event.getSource() == i7 ){ try{ fileLoader = new FileDialog( mainFrame, "Browse", FileDialog.LOAD ); fileLoader.show(); loadFile( fileLoader.getFile(), "i7" ); } catch( SecurityException se6 ){ errorArea.setText( "Sorry Could open file dialog box. You may want to copy the\nfeature vectors for testing and paste it in the text area.\n" + se6 ); } } else if( event.getSource() == i7a ){ try{ fileLoader = new FileDialog( mainFrame, "Browse", FileDialog.LOAD ); fileLoader.show(); loadFile( fileLoader.getFile(), "i7a" ); } catch( SecurityException se7 ){ errorArea.setText( "Sorry Could open file dialog box. You may want to copy the\ntarget vectors for testing and paste it in the text area.\n" + se7 ); } } else if( event.getSource() == i8 ){ targetFrame.setVisible( false ); } else if( event.getSource() == i9 ){ //Display help using targetFrame ( testing vectors ) testVecHelp(); } else if( event.getSource() == i12 ){ try{ fileLoader = new FileDialog( mainFrame, "Browse", FileDialog.LOAD ); fileLoader.show(); loadFile( fileLoader.getFile(), "i12" ); } catch( SecurityException se8 ){ errorArea.setText( "Sorry Could open file dialog box. You may want to copy the\nfeature vectors for validation and paste it in the text area.\n" + se8 ); } } else if( event.getSource() == i13 ){ try{ fileLoader = new FileDialog( mainFrame, "Browse", FileDialog.LOAD ); fileLoader.show(); loadFile( fileLoader.getFile(), "i13" ); } catch( SecurityException se9 ){ errorArea.setText( "Sorry Could open file dialog box. You may want to copy the\ntarget vectors for validation and paste it in the text area.\n" + se9 ); } } else if( event.getSource() == i14 ){ valFrame.setVisible( false ); } else if( event.getSource() == i15 ){ //Display help using valFrame ( validation vectors ) valVecHelp(); } else if( event.getSource() == Default_Param ) { tf1.setText( "" ); tf2.setText( "" ); tf3.setText( "" ); tf4.setText( "" ); sigmoid.setSelectedCheckbox( unipolar ); tf5.setText( "1.82" ); tf6.setText( "0.04" ); tf7.setText( "2.0" ); tf8.setText( "1.82" ); tf9.setText( "0.04" ); tf10.setText( "2.0" ); tf11.setText( "0.6" ); tf12.setText( "1.0" ); tf13.setText( "0.1" ); tf14.setText( "" ); } else if( event.getSource() == submit ){ try{ N = tf1.getText(); M = tf2.getText(); J = tf3.getText(); Q = tf4.getText(); a1 = tf5.getText(); da1 = tf6.getText(); b1 = tf7.getText(); a2 = tf8.getText(); da2 = tf9.getText(); b2 = tf10.getText(); e1 = tf11.getText(); e2 = tf12.getText(); mu = tf13.getText(); num = tf14.getText(); vN = Integer.valueOf( N ).intValue(); vM = Integer.valueOf( M ).intValue(); vJ = Integer.valueOf( J ).intValue(); vQ = Integer.valueOf( Q ).intValue(); vnum = Integer.valueOf( num ).intValue(); va1 = Double.valueOf( a1 ).doubleValue(); va2 = Double.valueOf( a2 ).doubleValue(); vda1 = Double.valueOf( da1 ).doubleValue(); vda2 = Double.valueOf( da2 ).doubleValue(); vb1 = (double)vN/( Double.valueOf( b1 ).doubleValue() ); vb2 = (double)vM/( Double.valueOf( b2 ).doubleValue() ); ve1 = Double.valueOf( e1 ).doubleValue(); ve2 = Double.valueOf( e2 ).doubleValue(); vmu = Double.valueOf( mu ).doubleValue(); //allocate arrays for weights W = new double[vN][vM]; delW = new double[vN][vM]; ndelW = new double[vN][vM]; U = new double[vM][vJ]; delU = new double[vM][vJ]; ndelW = new double[vM][vJ]; rSum = new double[vQ][vM]; sSum = new double[vQ][vJ]; Y = new double[vQ][vM]; Z = new double[vQ][vJ]; Err = new double[vnum]; temp = new double[vnum]; paramFrame.setVisible( false ); } catch ( NumberFormatException nfe ){ errorArea.setText( "Sorry, you have entered an incorrect input:\n" + nfe ); } try{ if( unipolar == sigmoid.getSelectedCheckbox() ){ } else if( bipolar == sigmoid.getSelectedCheckbox() ){ } else throw new CheckBoxNotSelectedException( "Radio button not selected" ); } catch( CheckBoxNotSelectedException cb1 ){ errorArea.setText( "You must select a sigmoid function:\n" + cb1 ); } } else if( event.getSource() == trainVec ){ featureFrame.setSize( 475, 575 ); featureFrame.setVisible( true ); } else if( event.getSource() == valVec ){ valFrame.setSize( 475, 575 ); valFrame.setVisible( true ); } else if( event.getSource() == submit1 ){ try{ if( ( vQ == 0 && vN == 0 && vJ == 0 ) ) throw new ZeroInitializedDataException( "Input data missing" ); featureVectors = featureArea.getText(); fv = new StringTokenizer( featureVectors ); X = new double[vQ][vN]; //Scan the input feature vectors for training for( int i = 0; i < vQ; i++ ){ for( int j = 0; j < vN; j++ ){ X[i][j] = Double.valueOf( fv.nextToken() ).doubleValue(); } } featureVectors1 = featureArea1.getText(); fv1 = new StringTokenizer( featureVectors1 ); T = new double[vQ][vJ]; //Scan the input target vectors for training for( int i = 0; i < vQ; i++ ){ for( int j = 0; j < vJ; j++ ){ T[i][j] = Double.valueOf( fv1.nextToken() ).doubleValue(); } } featureFrame.setVisible( false ); } catch( NoSuchElementException nse1 ){ errorArea.setText( "You must enter the input vectors for training:\n" + nse1 ); } catch( ZeroInitializedDataException zide5 ){ errorArea.setText( "You must either enter the input vectors for training:\nOr close the window and input the network parameters.:\n" + zide5 ); } catch( NumberFormatException nfe5 ){ errorArea.setText( "Sorry. Input format is incorrect:\n" + nfe5 ); } } else if( event.getSource() == testVec ){ targetFrame.setSize( 475, 575 ); targetFrame.setVisible( true ); } else if( event.getSource() == submit2 ){ try{ testQ = tf15.getText(); vtestQ = Integer.valueOf( testQ ).intValue(); if( vJ == 0 ) throw new NoSuchElementException(); targetVectors = targetArea.getText(); tv = new StringTokenizer( targetVectors ); testX = new double[vtestQ][vN]; //Scan the input feature vectors for testing for( int i = 0; i < vtestQ; i++ ){ for( int j = 0; j < vN; j++ ){ testX[i][j] = Double.valueOf( tv.nextToken() ).doubleValue(); } } targetVectors1 = targetArea1.getText(); tv1 = new StringTokenizer( targetVectors1 ); testT = new double[vtestQ][vJ]; //Scan the input target vectors for testing for( int i = 0; i < vtestQ; i++ ){ for( int j = 0; j < vJ; j++ ){ testT[i][j] = Double.valueOf( tv1.nextToken() ).doubleValue(); } } testrSum = new double[vtestQ][vM]; testsSum = new double[vtestQ][vJ]; testY = new double[vtestQ][vM]; testZ = new double[vtestQ][vJ]; targetFrame.setVisible( false ); } catch( NoSuchElementException nse2 ){ errorArea.setText( "You must enter the input vectors for testing:\nOr you have not entered all the network parameters:\n" + nse2 ); } catch( NumberFormatException nfe6 ){ errorArea.setText( "Sorry. Input format is incorrect:\n" + nfe6 ); } } else if( event.getSource() == train ){ try{ if( vnum == 0 ) throw new ZeroInitializedDataException( "No parameters entered to train the network" ); frozen = true; iteration = 0; start(); } catch( ZeroInitializedDataException zie3 ){ errorArea.setText( "You must enter all the network parameters to train the network:\n" + zie3 ); } } else if( event.getSource() == stopBtn ){ frozen = false; stop(); } else if( event.getSource() == clear ){ //Clear the error text area errorArea.setText( "" ); } else if( event.getSource() == errorStat ){ try{ double minErr = temp[0]; double maxErr = temp[vnum-1]; int minIte = 0; int maxIte = vnum - 1; boolean minFlag = true; boolean maxFlag = true; for( int i = 0; i < vnum; i++ ){ if( ( minErr == Err[i] ) && minFlag ){ minIte = i; minIte++; minFlag = false; } if( ( maxErr == Err[i] ) && maxFlag ){ maxIte = i; maxIte++; maxFlag = false; } } errorArea.append( "MAXIMUM Total Errors at Iteration = " ); errorArea.append( iteStr.valueOf( maxIte ) ); errorArea.append( ":\n" ); errorArea.append( "Sum-Squared Error " ); errorArea.append( errStr.valueOf( maxErr ) ); errorArea.append( "\n" ); errorArea.append( "Root Sum-Squared Error " ); errorArea.append( errStr.valueOf( Math.sqrt( maxErr ) ) ); errorArea.append( "\n" ); errorArea.append( "Mean-Squared Error " ); errorArea.append( errStr.valueOf( maxErr/(double)vQ ) ); errorArea.append( "\n" ); errorArea.append( "Root Mean-Squared Error " ); errorArea.append( errStr.valueOf( Math.sqrt( maxErr/(double)vQ ) ) ); errorArea.append( "\n\n" ); errorArea.append( "MINIMUM Total Errors at Iteration = " ); errorArea.append( iteStr.valueOf( minIte ) ); errorArea.append( ":\n" ); errorArea.append( "Sum-Squared Error " ); errorArea.append( errStr.valueOf( minErr ) ); errorArea.append( "\n" ); errorArea.append( "Root Sum-Squared Error " ); errorArea.append( errStr.valueOf( Math.sqrt( minErr ) ) ); errorArea.append( "\n" ); errorArea.append( "Mean-Squared Error " ); errorArea.append( errStr.valueOf( minErr/(double)vQ ) ); errorArea.append( "\n" ); errorArea.append( "Root Mean-Squared Error " ); errorArea.append( errStr.valueOf( Math.sqrt( minErr/(double)vQ ) ) ); errorArea.append( "\n\n" ); } catch( NullPointerException np2 ){ errorArea.setText( "There are no errors to report:\n" + np2 ); } } else if( event.getSource() == param ){ try{ if( (va1 == 0 ) || (va2 == 0 ) ) throw new ZeroInitializedDataException( "No learned parameters available" ); errorArea.append( "Exponential Rates:\n" ); errorArea.append( "alpha1: " ); errorArea.append( iteStr.valueOf( va1 ) ); errorArea.append( " alpha2: " ); errorArea.append( iteStr.valueOf( va2 ) ); errorArea.append( "\n\n" ); errorArea.append( "Learning Rates:\n" ); errorArea.append( "eta1: " ); errorArea.append( iteStr.valueOf( ve1 ) ); errorArea.append( " eta2: " ); errorArea.append( iteStr.valueOf( ve2 ) ); errorArea.append( "\n\n" ); if( unipolar == sigmoid.getSelectedCheckbox() ){ errorArea.append( "Biases:\n" ); errorArea.append( "bias1: " ); errorArea.append( iteStr.valueOf( vb1 ) ); errorArea.append( " bias2: " ); errorArea.append( iteStr.valueOf( vb2 ) ); errorArea.append( "\n\n" ); } errorArea.append( "Learned weights for middle neurodes ( Wnm ):\n" ); for( int i = 0; i < vN; i++ ){ for( int j = 0; j < vM; j++ ){ errorArea.append( iteStr.valueOf( W[i][j] ) ); errorArea.append( " " ); } errorArea.append( "\n" ); } errorArea.append( "\n" ); errorArea.append( "Learned weights for output neurodes ( Umj ):\n" ); for( int i = 0; i < vM; i++ ){ for( int j = 0; j < vJ; j++ ){ errorArea.append( iteStr.valueOf( U[i][j] ) ); errorArea.append( " " ); } errorArea.append( "\n" ); } } catch( ZeroInitializedDataException zid2 ){ errorArea.setText( "The network has no learned parameters at this time:\n" + zid2 ); } } else if( event.getSource() == test ){ try{ verErr = TestPredict_NN(); //Display target and predicted vectors errorArea.append( "Target output " ); errorArea.append( "Predicted output\n" ); errorArea.append( "----------------------------------------\n" ); for( int i = 0; i < vtestQ; i++ ){ for( int j = 0; j < vJ; j++ ){ errorArea.append( iteStr.valueOf( testT[i][j] ) ); errorArea.append( " " ); errorArea.append( iteStr.valueOf( testZ[i][j] ) ); errorArea.append( "\n" ); } errorArea.append( "----------------------------------------\n" ); } errorArea.append( "\n" ); errorArea.append( "Mean-Square Errors:\n" ); errorArea.append( "Error on the last training iteration " ); errorArea.append( iteStr.valueOf( Err[vnum-1]/(double)vQ ) ); errorArea.append( "\n" ); errorArea.append( "Error in the testing phase " ); errorArea.append( iteStr.valueOf( verErr/(double)vtestQ ) ); errorArea.append( "\n\n" ); if( ( verErr/(double)vtestQ ) < 2.0*( Err[vnum-1]/(double)vQ ) ) errorArea.append( "Training of the network is acceptable\n" ); else errorArea.append( "Training of the network is NOT acceptable\n" ); Zgrph = new double[vtestQ]; Tgrph = new double[vtestQ]; //Sort the temporary arrays Zgrph and Tgrph for( int i = 0; i < vtestQ; i++ ){ for( int j = 0; j < vJ; j++ ){ Zgrph[i] = testZ[i][j]; Tgrph[i] = testT[i][j]; } } QuickSort( Zgrph, 0, ( vtestQ - 1 ) ); QuickSort( Tgrph, 0, ( vtestQ - 1 ) ); G2Init(); } catch( NullPointerException np3 ){ errorArea.setText( "You must first train the network to test it:\n" + np3 ); } } else if( event.getSource() == submit3 ){ try{ valQ = tf16.getText(); vvalQ = Integer.valueOf( valQ ).intValue(); if( vJ == 0 ) throw new NoSuchElementException(); valFeatureVectors = valFeatureArea.getText(); val1 = new StringTokenizer( valFeatureVectors ); valX = new double[vvalQ][vN]; //Scan the input feature vectors for testing for( int i = 0; i < vvalQ; i++ ){ for( int j = 0; j < vN; j++ ){ valX[i][j] = Double.valueOf( val1.nextToken() ).doubleValue(); } } valTargetVectors = valTargetArea.getText(); val2 = new StringTokenizer( valTargetVectors ); valT = new double[vvalQ][vJ]; //Scan the input target vectors for testing for( int i = 0; i < vvalQ; i++ ){ for( int j = 0; j < vJ; j++ ){ valT[i][j] = Double.valueOf( val2.nextToken() ).doubleValue(); } } valrSum = new double[vvalQ][vM]; valsSum = new double[vvalQ][vJ]; valY = new double[vvalQ][vM]; valZ = new double[vvalQ][vJ]; valFrame.setVisible( false ); } catch( NoSuchElementException nse3 ){ errorArea.setText( "You must enter the input vectors for validation:\nOr you have not entered all the network parameters:\n" + nse3 ); } catch( NumberFormatException nfe7 ){ errorArea.setText( "Sorry. Input format is incorrect:\n" + nfe7 ); } } else if( event.getSource() == validate ){ try{ if( ( oldValErr == 0.0 ) || ( newValErr == 0.0 ) ) throw new ZeroInitializedDataException( "No validation error data available" ); if( valFlag ){ oldValErr = ValidatePredict_NN(); newValErr = oldValErr; valFlag = false; } else newValErr = ValidatePredict_NN(); //Display old and new validation errors errorArea.append( "Total Sum-Squared Errors for Validation:\n" ); errorArea.append( "Previous Error " ); errorArea.append( iteStr.valueOf( oldValErr ) ); errorArea.append( "\n" ); errorArea.append( "Current Error " ); errorArea.append( iteStr.valueOf( newValErr ) ); errorArea.append( "\n\n" ); if( newValErr <= oldValErr ) errorArea.append( "Must continue training the network\n" ); else errorArea.append( "The network is ready for testing\n" ); } catch( ZeroInitializedDataException zid1 ){ errorArea.setText( "You must first train the network in order to perform validation:\n" + zid1 ); } } else if( event.getSource() == graph1 ){ try{ if( vnum == 0 ) throw new NullPointerException(); gFrame1.setSize( 400, 450 ); gFrame1.setVisible( true ); } catch( NullPointerException np8 ){ errorArea.setText( "Data not available to draw the Error Graph:\n" + np8 ); } } else if( event.getSource() == close1 ){ gFrame1.setVisible( false ); } else if( event.getSource() == graph2 ){ try{ if( vtestQ == 0 ) throw new NullPointerException(); gFrame2.setSize( 400, 450 ); gFrame2.setVisible( true ); } catch( NullPointerException np5 ){ errorArea.setText( "Data not available to draw the Prediction Graph:\n" + np5 ); } } else if( event.getSource() == close2 ){ gFrame2.setVisible( false ); } }//END Event handling public void paint( Graphics g ) { } //NN training methods ---- public void InitializeWeights(){ //Local variables double beta; //initialize weights between 0.1 and 0.9 for unipolar sigmoid //initialize weights between -0.9 and 0.9 for bipolar sigmoid for( int i = 0; i < vN; i++ ){ for( int j = 0; j < vM; j++ ){ if( unipolar == sigmoid.getSelectedCheckbox() ) W[i][j] = ( Math.random() )*0.8 + 0.1; else W[i][j] = ( Math.random() )*1.8 - 0.9; //sum weights at hidden neurodes sumW = sumW + W[i][j]; } } for( int i = 0; i < vM; i++ ){ for( int j = 0; j < vJ; j++ ){ if( unipolar == sigmoid.getSelectedCheckbox() ) U[i][j] = ( Math.random() )*0.8 + 0.1; else U[i][j] = ( Math.random() )*1.8 - 0.9; //Russo adjustment of initial weights U[i][j] = U[i][j]*( 2.4/(double)vM ); } } //compute Nguyen-Widrow factor beta = ( 0.7*( Math.exp( (Math.log((double)vM))/((double)vN) ) ) )/sumW; //Nguyen-Widrow adjustment for( int i = 0; i < vN; i++ ){ for( int j = 0; j < vM; j++ ){ W[i][j] = beta*W[i][j]; } } //compute new SSE, MLP outputs oldErr = Update_NN(); //make the first time flag false //initFlag = false; } //END InitializeWeights() public void AdjustWeights(){ //Local variables double umjSum, wnmSum; for( int i = 0; i < vM; i++ ){ for( int j = 0; j < vJ; j++ ){ umjSum = 0.0; for( int k = 0; k < vQ; k++ ){ if( unipolar == sigmoid.getSelectedCheckbox() ) dgds = va2*Z[k][j]*( 1.0 - Z[k][j] ); else dgds = ( va2*( 1.0 + Z[k][j] )*( 1.0 - Z[k][j] ) )/2.0; if( dgds < 0.1 ) dgds = 0.1; umjSum = umjSum + ( T[k][j] - Z[k][j] )*dgds*Y[k][i]; } delU[i][j] = ve2*umjSum; U[i][j] = U[i][j] + delU[i][j]; } } newErr = Update_NN(); StepGain_Adjust( ve2, true ); for( int i = 0; i < vN; i++ ){ for( int j = 0; j < vM; j++ ){ wnmSum = 0.0; for( int k = 0; k < vQ; k++ ){ for( int l = 0; l < vJ; l++ ){ if( unipolar == sigmoid.getSelectedCheckbox() ) dhdr = va1*Y[k][j]*( 1.0 - Y[k][j] ); else dhdr = ( va1*( 1.0 + Y[k][j] )*( 1.0 - Y[k][j] ) )/2.0; if( dhdr < 0.1 ) dhdr = 0.1; wnmSum = wnmSum + ( T[k][l] - Z[k][l] )*dhdr*U[j][l]*dgds*X[k][i]; } } delW[i][j] = ve1*wnmSum; W[i][j] = W[i][j] + delW[i][j]; } } newErr = Update_NN(); StepGain_Adjust( ve1, false ); }//END AdjustWeights() public double Update_NN(){ for( int i = 0; i < vQ; i++ ){ for( int j = 0; j < vM; j++ ){ rSum[i][j] = 0.0; for( int k = 0; k < vN; k++ ){ rSum[i][j] = rSum[i][j] + W[k][j]*X[i][k]; } if( unipolar == sigmoid.getSelectedCheckbox() ) Y[i][j] = 1.0/( 1.0 + Math.exp( ( -va1*rSum[i][j] + vb1 ) ) ); else Y[i][j] = ( 1.0 + Math.exp( -va1*rSum[i][j] ) )/( 1.0 - Math.exp( -va1*rSum[i][j] ) ); } for( int l = 0; l < vJ; l++ ){ sSum[i][l] = 0.0; for( int m = 0; m < vM; m++ ){ sSum[i][l] = sSum[i][l] + U[m][l]*Y[i][m]; } if( unipolar == sigmoid.getSelectedCheckbox() ) Z[i][l] = 1.0/( 1.0 + Math.exp( ( -va2*sSum[i][l] + vb2 ) ) ); else Z[i][l] = ( 1.0 + Math.exp( -va2*sSum[i][l] ) )/( 1.0 - Math.exp( -va2*sSum[i][l] ) ); } } ssErr = 0.0; for( int i = 0; i < vQ; i++ ){ for( int j = 0; j < vJ; j++ ){ ssErr = ssErr + ( T[i][j] - Z[i][j] )*( T[i][j] - Z[i][j] ); } } return ssErr; }//END Update_NN() public void StepGain_Adjust( double eta, boolean key ){ if( newErr < oldErr ) eta = 1.024*eta; else eta = 0.92*eta; if( key ) ve2 = eta; else ve1 = eta; }//END StepGain_Adjust() public void AdjustAlphas(){ if( ( iteration%5 == 0 ) && ( iteration < vnum/2 ) ) { vah = va1; va1 = va1 + vda1; } newErr = Update_NN(); if( newErr < oldErr ) { oldErr = newErr; vda1 = vda1*1.1; } else { va1 = vah - vda1; } newErr = Update_NN(); if( newErr < oldErr ) { oldErr = newErr; vda1 = vda1*1.1; } else { va1 = vah; vda1 = vda1*0.9; } vah = va2; va2 = va2 + vda2; newErr = Update_NN(); if( newErr < oldErr ) { oldErr = newErr; vda2 = vda2*1.1; } else { va2 = vah - vda2; } newErr = Update_NN(); if( newErr < oldErr ) { oldErr = newErr; vda2 = vda2*1.1; } else { va2 = vah; vda2 = vda2*0.9; } }//END AdjustAlphas() public void QuickSort( double temp1[], int p, int r ){ if( p < r ){ int q = Split( temp1, p, r ); QuickSort( temp1, p, q ); QuickSort( temp1, q+1, r ); } }//END QuickSort public int Split( double temp1[], int p, int r ){ double x = temp1[p]; int i = p - 1; int j = r + 1; while( i < j ){ j = j - 1; while( temp1[j] > x ){ j = j - 1; } i = i + 1; while( temp1[i] < x ){ i = i + 1; } if( i < j ){ double tmp = temp1[j]; temp1[j] = temp1[i]; temp1[i] = tmp; } else return j; } return i; }//END Split public void loadFile( String filename, String flag ) throws SecurityException{ try{ File file = new File( filename ); FileInputStream in = new FileInputStream( file ); int fileLen = (int)file.length(); byte[] fileContent = new byte[ fileLen ]; in.read( fileContent ); String fileContentStr = new String( fileContent ); if( flag.equals( "i4" ) ) featureArea.setText( fileContentStr ); else if( flag.equals( "i4a" ) ) featureArea1.setText( fileContentStr ); else if( flag.equals( "i7" ) ) targetArea.setText( fileContentStr ); else if( flag.equals( "i7a" ) ) targetArea1.setText( fileContentStr ); else if( flag.equals( "i12" ) ) valFeatureArea.setText( fileContentStr ); else valTargetArea.setText( fileContentStr ); } catch( IOException ioe1 ){ if( flag.equals( "i4" ) ) featureArea.setText( "File IOError: " + ioe1 ); else if( flag.equals( "i4a" ) ) featureArea1.setText( "File IOError: " + ioe1 ); else if( flag.equals( "i7" ) ) targetArea.setText( "File IOError: " + ioe1 ); else if( flag.equals( "i7a" ) ) targetArea1.setText( "File IOError: " + ioe1 ); else if( flag.equals( "i12" ) ) valFeatureArea.setText( "File IOError: " + ioe1 ); else valTargetArea.setText( "File IOError: " + ioe1 ); } catch( SecurityException se1 ){ if( flag.equals( "i4" ) ) featureArea.setText( "Sorry Could not open file..." + filename + ": " + se1 ); else if( flag.equals( "i4a" ) ) featureArea1.setText( "Sorry Could not open file..." + filename + ": " + se1 ); else if( flag.equals( "i7" ) ) targetArea.setText( "Sorry Could not open file..." + filename + ": " + se1 ); else if( flag.equals( "i7a" ) ) valFeatureArea.setText( "Sorry Could not open file..." + filename + ": " + se1 ); else valTargetArea.setText( "Sorry Could not open file..." + filename + ": " + se1 ); } } public void saveFile( String filename ) throws SecurityException{ try{ File file = new File( filename ); FileOutputStream out = new FileOutputStream( file ); byte[] fileContent = errorArea.getText().getBytes(); out.write( fileContent ); } catch( IOException ioe2 ){ errorArea.setText( "File IOError: " + ioe2 ); } catch( SecurityException se2 ){ errorArea.setText( "Sorry Could not save file..." + filename + ": " + se2 ); } } public double TestPredict_NN(){ for( int i = 0; i < vtestQ; i++ ){ for( int j = 0; j < vM; j++ ){ testrSum[i][j] = 0.0; for( int k = 0; k < vN; k++ ){ testrSum[i][j] = testrSum[i][j] + W[k][j]*testX[i][k]; } if( unipolar == sigmoid.getSelectedCheckbox() ) testY[i][j] = 1.0/( 1.0 + Math.exp( ( -va1*testrSum[i][j] + vb1 ) ) ); else testY[i][j] = ( 1.0 + Math.exp( -va1*testrSum[i][j] ) )/( 1.0 - Math.exp( -va1*testrSum[i][j] ) ); } for( int l = 0; l < vJ; l++ ){ testsSum[i][l] = 0.0; for( int m = 0; m < vM; m++ ){ testsSum[i][l] = testsSum[i][l] + U[m][l]*testY[i][m]; } if( unipolar == sigmoid.getSelectedCheckbox() ) testZ[i][l] = 1.0/( 1.0 + Math.exp( ( -va2*testsSum[i][l] + vb2 ) ) ); else testZ[i][l] = ( 1.0 + Math.exp( -va2*testsSum[i][l] ) )/( 1.0 - Math.exp( -va2*testsSum[i][l] ) ); } } ssErr = 0.0; for( int i = 0; i < vtestQ; i++ ){ for( int j = 0; j < vJ; j++ ){ ssErr = ssErr + ( testT[i][j] - testZ[i][j] )*( testT[i][j] - testZ[i][j] ); } } return ssErr; }//END TestPredict_NN() public double ValidatePredict_NN(){ for( int i = 0; i < vvalQ; i++ ){ for( int j = 0; j < vM; j++ ){ valrSum[i][j] = 0.0; for( int k = 0; k < vN; k++ ){ valrSum[i][j] = valrSum[i][j] + W[k][j]*valX[i][k]; } if( unipolar == sigmoid.getSelectedCheckbox() ) valY[i][j] = 1.0/( 1.0 + Math.exp( ( -va1*valrSum[i][j] + vb1 ) ) ); else valY[i][j] = ( 1.0 + Math.exp( -va1*valrSum[i][j] ) )/( 1.0 - Math.exp( -va1*valrSum[i][j] ) ); } for( int l = 0; l < vJ; l++ ){ valsSum[i][l] = 0.0; for( int m = 0; m < vM; m++ ){ valsSum[i][l] = valsSum[i][l] + U[m][l]*valY[i][m]; } if( unipolar == sigmoid.getSelectedCheckbox() ) valZ[i][l] = 1.0/( 1.0 + Math.exp( ( -va2*valsSum[i][l] + vb2 ) ) ); else valZ[i][l] = ( 1.0 + Math.exp( -va2*valsSum[i][l] ) )/( 1.0 - Math.exp( -va2*valsSum[i][l] ) ); } } ssErr = 0.0; for( int i = 0; i < vvalQ; i++ ){ for( int j = 0; j < vJ; j++ ){ ssErr = ssErr + ( valT[i][j] - valZ[i][j] )*( valT[i][j] - valZ[i][j] ); } } return ssErr; }//END ValidatePredict_NN() //Thread Methods public void start(){ if(frozen){ //Start the loop! if (myThread == null) { myThread = new Thread(this); } myThread.start(); } }//end start public void stop(){ try{ if( myThread == null && vnum == 0 ) throw new NoThreadsStartedException( "Network training is not in progress" ); //Stop iterating. myThread = null; errorArea.append( "\nStop. Training...Done.\n" ); } catch( NoThreadsStartedException nts1 ){ errorArea.setText( "No threads are currently running:\n" + nts1 ); } } public void run(){ //Just to be nice, lower this thread's priority //so it can't interfere with other processing going on. Thread.currentThread().setPriority(Thread.MIN_PRIORITY); //Remember which thread we are. Thread currentThread = Thread.currentThread(); //This is the iterative loop. errorArea.append( "\nStart trainining...\n" ); Date begin = new Date(); InitializeWeights(); while (currentThread == myThread) { //Call NN training methods here //Adjust weights via Newton's linear approximation method AdjustWeights(); //Update the dacay rate parameter in the sigmoids AdjustAlphas(); if( iteration >= vnum ){ //Copy error array to temp array for( int i = 0; i < vnum; i++ ) temp[i] = Err[i]; //Sort the error array QuickSort( temp, 0, ( vnum - 1 ) ); G1Init(); stop(); break; } else Err[iteration++] = oldErr; if( iteration == ( vnum/2 ) ) vmu = 0.0; if( iteration%10 == 0 ){ if( iteration == 10 ){ errorArea.append( "Iteration" ); errorArea.append( " " ); errorArea.append( "Total Sum-Squared Error" ); errorArea.append( "\n" ); } errorArea.append( iteStr.valueOf( iteration ) ); errorArea.append( " " ); errorArea.append( errStr.valueOf( oldErr ) ); errorArea.append( "\n" ); } } Date end = new Date(); errorArea.append( "\nTotal time taken for training: " ); errorArea.append( errStr.valueOf( (end.getTime() - begin.getTime())/1000.00 ) ); errorArea.append( " seconds\n" ); }//end run public void G1Init(){ //Get min and max x-axis( iterations )/y-axis ( error ) g1xmin = 0.0; g1xmax = (double)( vnum ); g1ymin = temp[0]; g1ymax = temp[vnum-1]; //determine x-axis tick marks g1xtic1 = g1xmax/2.0; g1xtic2 = g1xtic1/2.0; g1xtic3 = g1xtic1 + g1xtic2; g1xtic4 = g1xmax; //determine y-axis tick marks g1ytic1 = g1ymax/2.0; g1ytic2 = g1ytic1/2.0; g1ytic3 = g1ytic1 + g1ytic2; g1ytic4 = g1ymax; } public void G2Init(){ //Get min and max x-axis( iterations )/y-axis ( error ) g2xmin = 0.0; g2xmax = (double)( vtestQ ); g2ymin1 = Zgrph[0]; g2ymax1 = Zgrph[vtestQ-1]; g2ymin2 = Tgrph[0]; g2ymax2 = Tgrph[vtestQ-1]; //determine x-axis tick marks g2xtic1 = g2xmax/2.0; g2xtic2 = g2xtic1/2.0; g2xtic3 = g2xtic1 + g2xtic2; g2xtic4 = g2xmax; } //To draw the error graph public class Graph1 extends Canvas{ public Graph1( Color bg, int width, int height ){ setBackground( bg ); setSize( width, height ); } public void paint( Graphics g ){ //x and y axis lines drawLine( g, 70, 50, 70, 250, Color.black ); drawLine( g, 70, 250, 270, 250, Color.black ); //draw x-axis tics and labels g.drawString( "|", tX( g1xtic1 ), 255 ); g.drawString( "|", tX( g1xtic2 ), 255 ); g.drawString( "|", tX( g1xtic3 ), 255 ); g.drawString( "|", tX( g1xtic4 ), 255 ); g.drawString( g1str1.valueOf( (int)g1xtic1 ), tX( g1xtic1 ) - 10, 265 ); g.drawString( g1str1.valueOf( (int)g1xtic2 ), tX( g1xtic2 ) - 10, 265 ); g.drawString( g1str1.valueOf( (int)g1xtic3 ), tX( g1xtic3 ) - 10, 265 ); g.drawString( g1str1.valueOf( (int)g1xtic4 ), tX( g1xtic4 ) - 10, 265 ); g.drawString( "Iterations", 160, 300 ); //draw y-axis tics and labels g.drawString( "-", 67, tY( g1ytic1 ) ); g.drawString( "-", 67, tY( g1ytic2 ) ); g.drawString( "-", 67, tY( g1ytic3 ) ); g.drawString( "-", 67, tY( g1ytic4 ) ); g.drawString( g1str1.valueOf( g1ytic1 ).substring(0,6), 25, tY( g1ytic1 ) ); g.drawString( g1str1.valueOf( g1ytic2 ).substring(0,6), 25, tY( g1ytic2 ) ); g.drawString( g1str1.valueOf( g1ytic3 ).substring(0,6), 25, tY( g1ytic3 ) ); g.drawString( g1str1.valueOf( g1ytic4 ).substring(0,6), 25, tY( g1ytic4 ) ); g.drawString( "Error", 45, 30 ); //draw the graph g1lastx = tX( g1xmin ); g1lasty = tY( Err[0] ); for( int i = 1; i < vnum; i++ ){ g1px1 = g1lastx; g1py1 = g1lasty; g1px2 = tX( (double)i ); g1py2 = tY( Err[i] ); drawLine( g, g1px1, g1py1, g1px2, g1py2, Color.red ); g1lastx = g1px2; g1lasty = g1py2; } } public void drawLine( Graphics g, int x1, int y1, int x2, int y2, Color c ){ Color origColor = g.getColor(); g.setColor( c ); g.drawLine( x1, y1, x2, y2 ); g.setColor( origColor ); } public int tX( double val1 ){ return (int)( 200.00*( val1 - g1xmin )/( g1xmax - g1xmin ) + 70.00 ); } public int tY( double val2 ){ return (int)( -200.00*( val2 - g1ymin )/( g1ymax - g1ymin ) + 250.00 ); } }//END Graph1 //To draw the prediction graph public class Graph2 extends Canvas{ public Graph2( Color bg, int width, int height ){ setBackground( bg ); setSize( width, height ); } public void paint( Graphics g ){ if( g2ymax1 > g2ymax2 ) g2ymax = g2ymax1; else g2ymax = g2ymax2; if( g2ymin1 < g2ymin2 ) g2ymin = g2ymin1; else g2ymin = g2ymin2; //determine y-axis tick marks g2ytic1 = g2ymax/2.0; g2ytic2 = g2ytic1/2.0; g2ytic3 = g2ytic1 + g2ytic2; g2ytic4 = g2ymax; //x and y axis lines drawLine( g, 70, 50, 70, 250, Color.black ); drawLine( g, 70, 250, 270, 250, Color.black ); //Legends drawLine( g, 50, 325, 100, 325, Color.blue ); g.drawString( "Target", 105, 325 ); drawLine( g, 200, 325, 250, 325, Color.red ); g.drawString( "Predicted", 255, 325 ); //draw x-axis tics and labels g.drawString( "|", tX( g2xtic1 ), 255 ); g.drawString( "|", tX( g2xtic2 ), 255 ); g.drawString( "|", tX( g2xtic3 ), 255 ); g.drawString( "|", tX( g2xtic4 ), 255 ); g.drawString( g2str1.valueOf( g2xtic1 ), tX( g2xtic1 ) - 7, 265 ); g.drawString( g2str1.valueOf( g2xtic2 ), tX( g2xtic2 ) - 7, 265 ); g.drawString( g2str1.valueOf( g2xtic3 ), tX( g2xtic3 ) - 7, 265 ); g.drawString( g2str1.valueOf( g2xtic4 ), tX( g2xtic4 ) - 7, 265 ); g.drawString( "Number of Exemplar Vectors", 120, 295 ); //draw y-axis tics and labels g.drawString( "-", 67, tY( g2ytic1 ) ); g.drawString( "-", 67, tY( g2ytic2 ) ); g.drawString( "-", 67, tY( g2ytic3 ) ); g.drawString( "-", 67, tY( g2ytic4 ) ); if( g2str1.valueOf( g2ytic1 ).length() > 4 ) g.drawString( g2str1.valueOf( g2ytic1 ).substring(0, 3), 45, tY( g2ytic1 ) ); else g.drawString( g2str1.valueOf( g2ytic1 ), 45, tY( g2ytic1 ) ); if( g2str1.valueOf( g2ytic2 ).length() > 4 ) g.drawString( g2str1.valueOf( g2ytic2 ).substring(0, 3), 45, tY( g2ytic2 ) ); else g.drawString( g2str1.valueOf( g2ytic2 ), 45, tY( g2ytic2 ) ); if( g2str1.valueOf( g2ytic3 ).length() > 4 ) g.drawString( g2str1.valueOf( g2ytic3 ).substring(0, 3), 45, tY( g2ytic3 ) ); else g.drawString( g2str1.valueOf( g2ytic3 ), 45, tY( g2ytic3 ) ); if( g2str1.valueOf( g2ytic4 ).length() > 4 ) g.drawString( g2str1.valueOf( g2ytic4 ).substring(0, 3), 45, tY( g2ytic4 ) ); else g.drawString( g2str1.valueOf( g2ytic4 ), 45, tY( g2ytic4 ) ); g.drawString( "Output", 45, 30 ); //draw the graph g2lastx = tX( g2xmin ); g2lasty = tY( Zgrph[0] ); for( int i = 1; i < vtestQ; i++ ){ g2px1 = g2lastx; g2py1 = g2lasty; g2px2 = tX( (double)i ); g2py2 = tY( Zgrph[i] ); drawLine( g, g2px1, g2py1, g2px2, g2py2, Color.red ); g2lastx = g2px2; g2lasty = g2py2; } g2lastx = tX( g2xmin ); g2lasty = tY( Tgrph[0] ); for( int i = 1; i < vtestQ; i++ ){ g2px1 = g2lastx; g2py1 = g2lasty; g2px2 = tX( (double)i ); g2py2 = tY( Tgrph[i] ); drawLine( g, g2px1, g2py1, g2px2, g2py2, Color.blue ); g2lastx = g2px2; g2lasty = g2py2; } } public void drawLine( Graphics g, int x1, int y1, int x2, int y2, Color c ){ Color origColor = g.getColor(); g.setColor( c ); g.drawLine( x1, y1, x2, y2 ); g.setColor( origColor ); } public int tX( double val1 ){ return (int)( 200.00*( val1 - g2xmin )/( g2xmax - g2xmin ) + 70.00 ); } public int tY( double val2 ){ return (int)( -200.00*( val2 - g2ymin )/( g2ymax - g2ymin ) + 250.00 ); } }//END Graph2 public void MainHelp(){ errorArea.append( "Multiple-Layered Perceptrons( MLPs ) Applet Help:\n\n" ); errorArea.append( "Multiple-layered perceptrons are feedforward artificial\n" ); errorArea.append( "neural networks widely used for pattern classification.\n" ); errorArea.append( "This applet uses the method of cross-validation in which\n" ); errorArea.append( "the data set is broken into three disjoint sets: a training\n" ); errorArea.append( "set, a validation set and a testing set. Network's performance\n" ); errorArea.append( "on the validation set is used to determine when to stop training.\n" ); errorArea.append( "This applet uses the full propagation algorithm to train the\n" ); errorArea.append( "network.\n\n" ); errorArea.append( "Please follow these steps to get going!!:\n\n" ); errorArea.append( "1. Click on the 'Input MLPs Parameters' button to\n" ); errorArea.append( " input all the preliminary network parameters.\n\n" ); errorArea.append( "2. Click on the 'Input Vectors for Training' button to\n" ); errorArea.append( " input all the feature and target vectors for training.\n\n" ); errorArea.append( "3. Click on the 'Input Vectors for Validation' button to\n" ); errorArea.append( " input all the feature and target vectors for validation.\n\n" ); errorArea.append( "4. Click on the 'Input Vectors for Testing' button to\n" ); errorArea.append( " input all the feature and target vectors for testing.\n\n" ); errorArea.append( "5. Now you are ready to train, validate, and test the network.\n" ); errorArea.append( " To train the network click on the 'Start Training' button.\n" ); errorArea.append( " While the network is being trained, Total Sum-Squared Error\n" ); errorArea.append( " at every 10 iterations is displayed in this text area. You can\n" ); errorArea.append( " stop training the network anytime by pressing the 'Stop' button.\n" ); errorArea.append( " The training will automatically stop when number of iterations are\n" ); errorArea.append( " exhausted.\n\n" ); errorArea.append( "6. Click on the 'View Error Graph' button to graphically visualize\n" ); errorArea.append( " the errors. Press 'Close' button to quit the graph.\n\n" ); errorArea.append( "7. Click on the 'View Error Statistics' button to view total minimum\n" ); errorArea.append( " and maximum Sum-Squared Error, Root Sum-Squared Error,\n" ); errorArea.append( " Mean-Square error, and Root Mean-Squared Error.\n\n" ); errorArea.append( "8. To Validate the network press 'Validate NN' button.\n" ); errorArea.append( " The previous and current Total Mean-Square Errors are displayed\n" ); errorArea.append( " in this text area. If the current error is greater than the previous\n" ); errorArea.append( " error then go on with the testing else repeat the training and the\n" ); errorArea.append( " the validation steps.\n\n" ); errorArea.append( "9. Once the network is cross-validated, click on the 'Start Testing'\n" ); errorArea.append( " button to test the network performance. The target and the predicted\n" ); errorArea.append( " network output is displayed in this text area. To visualize the results\n" ); errorArea.append( " graphically, click on the 'View Prediction Graph' button.\n\n" ); errorArea.append( "10. Click on the 'View Learned Parameters' to view exponential rates,\n" ); errorArea.append( " learning rates, biases, and the weights at the hidden and the output\n" ); errorArea.append( " neurodes.\n\n" ); errorArea.append( "11. To save the contents of the text area, select 'Save' option from the 'File'\n" ); errorArea.append( " menu. Due to some serious security considerations, some browsers\n" ); errorArea.append( " may not be able to save the contents. However, you may copy or cut\n" ); errorArea.append( " the contents and paste it in your Word or Notepad application file.\n\n" ); errorArea.append( "12. All the exceptions are caught and displayed in this text area\n\n" ); errorArea.append( "13. To exit the applet, select 'Quit' option from the 'File' menu.\n\n" ); errorArea.append( "14. To launch the applet again, press 'Reload' button on your browser.\n\n" ); errorArea.append( "15. To clear this text area, click on the 'Refresh' button.\n\n" ); } public void paramFrameHelp(){ helpArea.append( "1. First click on the 'View Default Parameters' button to\n" ); helpArea.append( " view default sigmoid function, exponential rates and the\n" ); helpArea.append( " increments, learning rates, biases, and momentum parameters.\n\n" ); helpArea.append( "2. Enter number of input nodes, output layer neurodes, middle\n" ); helpArea.append( " layer neurodes, exemplar vector, and iterations in the\n" ); helpArea.append( " respective text fields.\n\n" ); helpArea.append( "3. Select one of the sigmoid functions. The defaul is unipolar\n" ); helpArea.append( " sigmoid. The biases are ignored when bipolar sigmoid is selected.\n\n" ); helpArea.append( "4. Click on the 'Submit' button to input the parameters.\n" ); helpArea.append( " If any of the text field is empty, the NumberFormatException\n" ); helpArea.append( " is thrown and displayed in the main applet's text area.\n\n" ); helpArea.append( "5. After submitting the parameters, proceed to enter feature and\n" ); helpArea.append( " target vectors for training, validation, and testing.\n\n" ); helpArea.append( "6. To quit the 'MLPs Parameters' window, select 'Close' option\n" ); helpArea.append( " from the 'File' menu.\n\n" ); helpArea.append( "7. To quit the help window, click on the 'Close' button.\n" ); } public void trainVecHelp(){ featureArea.append( "Input Vectors for Training Help:\n\n" ); featureArea.append( "1. First select 'Feature Vectors' option from the 'File' menu.\n" ); featureArea.append( " A file dialog box will be displayed on your screen. Select\n" ); featureArea.append( " the file. The contents will be displayed in the first text\n" ); featureArea.append( " area.\n\n" ); featureArea.append( "2. Then select 'Target Vectors' option from the 'File' menu.\n" ); featureArea.append( " A file dialog box will be displayed on your screen. Select\n" ); featureArea.append( " the file. The contents will be displayed in the second text\n" ); featureArea.append( " area.\n\n" ); featureArea.append( "3. On some browsers you may not be able to bring up the file\n" ); featureArea.append( " dialog box for securiry reasons. In this case just copy the\n" ); featureArea.append( " data from your file and paste it in the respective text areas.\n\n" ); featureArea.append( "4. Click on 'Submit Vectors for Training' button to input feature and\n" ); featureArea.append( " target vectors.\n\n" ); featureArea.append( "5. Click on 'Clear' button to clear the contents of the text areas.\n\n" ); featureArea.append( "6. To quit the 'Input Vectors fot Training' window, select 'Close'\n" ); featureArea.append( " option from the 'File' menu.\n\n" ); } public void testVecHelp(){ targetArea.append( "Input Vectors for Testing Help:\n\n" ); targetArea.append( "1. Enter the number of exemplar vectors ( Q ) for testing\n" ); targetArea.append( " in the text field.\n\n" ); targetArea.append( "2. First select 'Feature Vectors' option from the 'File' menu.\n" ); targetArea.append( " A file dialog box will be displayed on your screen. Select\n" ); targetArea.append( " the file. The contents will be displayed in the first text\n" ); targetArea.append( " area.\n\n" ); targetArea.append( "3. Then select 'Target Vectors' option from the 'File' menu.\n" ); targetArea.append( " A file dialog box will be displayed on your screen. Select\n" ); targetArea.append( " the file. The contents will be displayed in the second text\n" ); targetArea.append( " area.\n\n" ); targetArea.append( "4. On some browsers you may not be able to bring up the file\n" ); targetArea.append( " dialog box for securiry reasons. In this case just copy the\n" ); targetArea.append( " data from your file and paste it in the respective text areas.\n\n" ); targetArea.append( "5. Click on 'Submit Vectors for Testing' button to input feature and\n" ); targetArea.append( " target vectors.\n\n" ); targetArea.append( "6. Click on 'Clear' button to clear the contents of the text areas.\n\n" ); targetArea.append( "7. To quit the 'Input Vectors for Testing' window, select 'Close'\n" ); targetArea.append( " option from the 'File' menu.\n\n" ); } public void valVecHelp(){ valFeatureArea.append( "Input Vectors for Validation Help:\n\n" ); valFeatureArea.append( "1. Enter the number of exemplar vectors ( Q ) for validation\n" ); valFeatureArea.append( " in the text field.\n\n" ); valFeatureArea.append( "2. First select 'Feature Vectors' option from the 'File' menu.\n" ); valFeatureArea.append( " A file dialog box will be displayed on your screen. Select\n" ); valFeatureArea.append( " the file. The contents will be displayed in the first text\n" ); valFeatureArea.append( " area.\n\n" ); valFeatureArea.append( "3. Then select 'Target Vectors' option from the 'File' menu.\n" ); valFeatureArea.append( " A file dialog box will be displayed on your screen. Select\n" ); valFeatureArea.append( " the file. The contents will be displayed in the second text\n" ); valFeatureArea.append( " area.\n\n" ); valFeatureArea.append( "4. On some browsers you may not be able to bring up the file\n" ); valFeatureArea.append( " dialog box for securiry reasons. In this case just copy the\n" ); valFeatureArea.append( " data from your file and paste it in the respective text areas.\n\n" ); valFeatureArea.append( "5. Click on 'Submit Vectors for Validation' button to input feature and\n" ); valFeatureArea.append( " target vectors.\n\n" ); valFeatureArea.append( "6. Click on 'Clear' button to clear the contents of the text areas.\n\n" ); valFeatureArea.append( "7. To quit the 'Input Vectors for Validation' window, select 'Close'\n" ); valFeatureArea.append( " option from the 'File' menu.\n\n" ); } //handle radio button exception public class CheckBoxNotSelectedException extends Exception{ public CheckBoxNotSelectedException(String message){ super(message); } } //handle uninitialized data exception public class ZeroInitializedDataException extends Exception{ public ZeroInitializedDataException(String message){ super(message); } } //handle null thread exception public class NoThreadsStartedException extends Exception{ public NoThreadsStartedException(String message){ super(message); } } }//END main applet