/*********************************************************************** This file is part of KEEL-software, the Data Mining tool for regression, classification, clustering, pattern mining and so on. Copyright (C) 2004-2010 F. Herrera (herrera@decsai.ugr.es) L. S�nchez (luciano@uniovi.es) J. Alcal�-Fdez (jalcala@decsai.ugr.es) S. Garc�a (sglopez@ujaen.es) A. Fern�ndez (alberto.fernandez@ujaen.es) J. Luengo (julianlm@decsai.ugr.es) This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/ **********************************************************************/ /** * <p> * @author Written by Luciano Sanchez (University of Oviedo) 01/01/2004 * @author Modified by Jose Otero (University of Oviedo) 01/12/2008 * @version 1.0 * @since JDK1.5 * </p> */ package keel.Algorithms.Statistical_Classifiers.ClassifierLinearLMS; import keel.Algorithms.Shared.Parsing.*; import keel.Algorithms.Statistical_Classifiers.Shared.DiscrAnalysis.*; import keel.Algorithms.Statistical_Classifiers.Shared.MatrixCalcs.*; import keel.Algorithms.Shared.ClassicalOptim.*; import org.core.*; import java.io.*; import java.util.StringTokenizer; import java.util.Vector; public class ClassifierLinearLMS { /** * <p> * In this class, a linear classifier using Least Mean Squares is implemented * </p> */ static Randomize rand; /** * <p> * In this method, a linear classifier using Least Mean Squares is estimated * @param tty unused boolean parameter, kept for compatibility * @param pc {@link ProcessConfig} object to obtain the train and test datasets * and the method's parameters. * </p> */ private static void linearClassifierLMS(boolean tty, ProcessConfig pc) { try { String line; ProcessDataset pd=new ProcessDataset(); line=(String)pc.parInputData.get(ProcessConfig.IndexTrain); if (pc.parNewFormat) pd.processClassifierDataset(line,true); else pd.oldClusteringProcess(line); int nData=pd.getNdata(); // Number of examples int nVariables=pd.getNvariables(); // Number of variables int nInputs=pd.getNinputs(); // Number of inputs pd.showDatasetStatistics(); double[][] X = pd.getX(); // Input data int[] C = pd.getC(); // Output data int [] Ct=new int[C.length]; int nClasses = pd.getNclasses(); // Number of classes double[] maxInput = pd.getImaximum(); // Maximum and minimum for input data double[] minInput = pd.getIminimum(); int[] nInputFolds=new int[nInputs]; // A vector is generated with classes 1 bit between n codified double Cbin[][] = new double[nData][nClasses]; for (int i=0;i<nData;i++) Cbin[i][C[i]]=1; for (int i=0;i<X.length;i++) Ct[i]=-1; // 1-layer perceptron int nLayers=0; int []ELEM=new int[nLayers]; // Weight vector (return value) int dimWeight=0; if (nLayers==0) { dimWeight=(nInputs+1)*(nClasses); } else { dimWeight=(nInputs+1)*ELEM[0]; for (int i=1;i<nLayers;i++) dimWeight+=(ELEM[i-1]+1)*(ELEM[i]); dimWeight+=(nClasses)*(ELEM[nLayers-1]+1); } double []weights=new double[dimWeight]; GCNet gcn=new GCNet(); double error=gcn.nntrain(nInputs,nClasses,X,Cbin,ELEM,weights,rand); double faults=0; double debugRMS=0; try { for (int i=0;i<X.length;i++) { double[] resp=gcn.nnoutput(X[i]); int theClass=AD.argmax(resp); for (int i1=0;i1<resp.length;i1++) debugRMS+=(resp[i1]-Cbin[i][i1])*(resp[i1]-Cbin[i][i1]); if (theClass!=C[i]) faults++; Ct[i]=theClass; } debugRMS/=X.length; System.out.println("Failures="+faults+" size="+nData); System.out.println("Debug RMS="+debugRMS); faults/=nData; System.out.println("Train error="+faults); } catch (Exception e) { System.out.println(e.toString()); } pc.trainingResults(C,Ct); // Test ProcessDataset pdt = new ProcessDataset(); int nTest,npInputs,npVariables; line=(String)pc.parInputData.get(ProcessConfig.IndexTest); if (pc.parNewFormat) pdt.processClassifierDataset(line,false); else pdt.oldClusteringProcess(line); nTest = pdt.getNdata(); npVariables = pdt.getNvariables(); npInputs = pdt.getNinputs(); pdt.showDatasetStatistics(); if (npInputs!=nInputs) throw new IOException("IOErr test file"); double[][] Xp=pdt.getX(); int [] Cp=pdt.getC(); int [] Co=new int[Cp.length]; // Test error try { faults=0; for (int i=0;i<Xp.length;i++) { double[] resp=gcn.nnoutput(Xp[i]); int theClass=AD.argmax(resp); if (theClass!=Cp[i]) faults++; Co[i]=theClass; } System.out.println("Failures="+faults+" total="+Xp.length); faults/=Cp.length; System.out.println("Test error="+faults); } catch (Exception e) { System.out.println(e.toString()); } pc.results(Cp,Co); // Now using the pseudoinverse matrix double[][]Ys = new double[X.length][nClasses]; double[][]Xs = new double[X.length][X[0].length+1]; double[][] A = new double[X[0].length+1][1]; for (int i=0;i<Xs.length;i++) { for (int j=0;j<nClasses;j++) Ys[i][j]=Cbin[i][j]; Xs[i][0]=1; for (int j=1;j<Xs[0].length;j++) Xs[i][j]=X[i][j-1]; } try { A=MatrixCalcs.matmul( MatrixCalcs.matmul( MatrixCalcs.inv( MatrixCalcs.matmul(MatrixCalcs.tr(Xs), Xs ) ), MatrixCalcs.tr(Xs) ),Ys); // Train error double Cs[][] = new double[Ys.length][Ys[0].length]; Cs = MatrixCalcs.matmul(Xs,A); debugRMS=0; for (int i=0;i<Cs.length;i++) for (int j=0;j<Cs[i].length;j++) debugRMS+=(Cs[i][j]-Ys[i][j])*(Cs[i][j]-Ys[i][j]); debugRMS/=Cs.length; System.out.println("DEBUG RMS PSEUDOINVERSE: "+debugRMS); } catch(Exception e) { System.err.println(e+" Matrix Calcs"); } } catch(FileNotFoundException e) { System.err.println(e+" Input file not found"); } catch(IOException e) { System.err.println(e+" Read error"); } } /** * <p> * This method runs {@link ClassifierLinearLMS} * @param args A vector of string with command line arguments * </p> */ public static void main(String args[]) { boolean tty=false; ProcessConfig pc=new ProcessConfig(); System.out.println("Reading configuration file: "+args[0]); if (pc.fileProcess(args[0])<0) return; int algo=pc.parAlgorithmType; rand=new Randomize(); rand.setSeed(pc.parSeed); ClassifierLinearLMS cl=new ClassifierLinearLMS(); cl.linearClassifierLMS(tty,pc); } }