[4] | 1 | /* |
---|
| 2 | * This program is free software; you can redistribute it and/or modify |
---|
| 3 | * it under the terms of the GNU General Public License as published by |
---|
| 4 | * the Free Software Foundation; either version 2 of the License, or |
---|
| 5 | * (at your option) any later version. |
---|
| 6 | * |
---|
| 7 | * This program is distributed in the hope that it will be useful, |
---|
| 8 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
| 9 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
| 10 | * GNU General Public License for more details. |
---|
| 11 | * |
---|
| 12 | * You should have received a copy of the GNU General Public License |
---|
| 13 | * along with this program; if not, write to the Free Software |
---|
| 14 | * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
---|
| 15 | */ |
---|
| 16 | |
---|
| 17 | /* |
---|
| 18 | * MIBoost.java |
---|
| 19 | * Copyright (C) 2005 University of Waikato, Hamilton, New Zealand |
---|
| 20 | * |
---|
| 21 | */ |
---|
| 22 | |
---|
| 23 | package weka.classifiers.mi; |
---|
| 24 | |
---|
| 25 | import weka.classifiers.Classifier; |
---|
| 26 | import weka.classifiers.AbstractClassifier; |
---|
| 27 | import weka.classifiers.SingleClassifierEnhancer; |
---|
| 28 | import weka.core.Capabilities; |
---|
| 29 | import weka.core.Instance; |
---|
| 30 | import weka.core.Instances; |
---|
| 31 | import weka.core.MultiInstanceCapabilitiesHandler; |
---|
| 32 | import weka.core.Optimization; |
---|
| 33 | import weka.core.Option; |
---|
| 34 | import weka.core.OptionHandler; |
---|
| 35 | import weka.core.RevisionUtils; |
---|
| 36 | import weka.core.TechnicalInformation; |
---|
| 37 | import weka.core.TechnicalInformationHandler; |
---|
| 38 | import weka.core.Utils; |
---|
| 39 | import weka.core.WeightedInstancesHandler; |
---|
| 40 | import weka.core.Capabilities.Capability; |
---|
| 41 | import weka.core.TechnicalInformation.Field; |
---|
| 42 | import weka.core.TechnicalInformation.Type; |
---|
| 43 | import weka.filters.Filter; |
---|
| 44 | import weka.filters.unsupervised.attribute.Discretize; |
---|
| 45 | import weka.filters.unsupervised.attribute.MultiInstanceToPropositional; |
---|
| 46 | |
---|
| 47 | import java.util.Enumeration; |
---|
| 48 | import java.util.Vector; |
---|
| 49 | |
---|
| 50 | /** |
---|
| 51 | <!-- globalinfo-start --> |
---|
| 52 | * MI AdaBoost method, considers the geometric mean of posterior of instances inside a bag (arithmatic mean of log-posterior) and the expectation for a bag is taken inside the loss function.<br/> |
---|
| 53 | * <br/> |
---|
| 54 | * For more information about Adaboost, see:<br/> |
---|
| 55 | * <br/> |
---|
| 56 | * Yoav Freund, Robert E. Schapire: Experiments with a new boosting algorithm. In: Thirteenth International Conference on Machine Learning, San Francisco, 148-156, 1996. |
---|
| 57 | * <p/> |
---|
| 58 | <!-- globalinfo-end --> |
---|
| 59 | * |
---|
| 60 | <!-- technical-bibtex-start --> |
---|
| 61 | * BibTeX: |
---|
| 62 | * <pre> |
---|
| 63 | * @inproceedings{Freund1996, |
---|
| 64 | * address = {San Francisco}, |
---|
| 65 | * author = {Yoav Freund and Robert E. Schapire}, |
---|
| 66 | * booktitle = {Thirteenth International Conference on Machine Learning}, |
---|
| 67 | * pages = {148-156}, |
---|
| 68 | * publisher = {Morgan Kaufmann}, |
---|
| 69 | * title = {Experiments with a new boosting algorithm}, |
---|
| 70 | * year = {1996} |
---|
| 71 | * } |
---|
| 72 | * </pre> |
---|
| 73 | * <p/> |
---|
| 74 | <!-- technical-bibtex-end --> |
---|
| 75 | * |
---|
| 76 | <!-- options-start --> |
---|
| 77 | * Valid options are: <p/> |
---|
| 78 | * |
---|
| 79 | * <pre> -D |
---|
| 80 | * Turn on debugging output.</pre> |
---|
| 81 | * |
---|
| 82 | * <pre> -B <num> |
---|
| 83 | * The number of bins in discretization |
---|
| 84 | * (default 0, no discretization)</pre> |
---|
| 85 | * |
---|
| 86 | * <pre> -R <num> |
---|
| 87 | * Maximum number of boost iterations. |
---|
| 88 | * (default 10)</pre> |
---|
| 89 | * |
---|
| 90 | * <pre> -W <class name> |
---|
| 91 | * Full name of classifier to boost. |
---|
| 92 | * eg: weka.classifiers.bayes.NaiveBayes</pre> |
---|
| 93 | * |
---|
| 94 | * <pre> -D |
---|
| 95 | * If set, classifier is run in debug mode and |
---|
| 96 | * may output additional info to the console</pre> |
---|
| 97 | * |
---|
| 98 | <!-- options-end --> |
---|
| 99 | * |
---|
| 100 | * @author Eibe Frank (eibe@cs.waikato.ac.nz) |
---|
| 101 | * @author Xin Xu (xx5@cs.waikato.ac.nz) |
---|
| 102 | * @version $Revision: 5928 $ |
---|
| 103 | */ |
---|
| 104 | public class MIBoost |
---|
| 105 | extends SingleClassifierEnhancer |
---|
| 106 | implements OptionHandler, MultiInstanceCapabilitiesHandler, |
---|
| 107 | TechnicalInformationHandler { |
---|
| 108 | |
---|
| 109 | /** for serialization */ |
---|
| 110 | static final long serialVersionUID = -3808427225599279539L; |
---|
| 111 | |
---|
| 112 | /** the models for the iterations */ |
---|
| 113 | protected Classifier[] m_Models; |
---|
| 114 | |
---|
| 115 | /** The number of the class labels */ |
---|
| 116 | protected int m_NumClasses; |
---|
| 117 | |
---|
| 118 | /** Class labels for each bag */ |
---|
| 119 | protected int[] m_Classes; |
---|
| 120 | |
---|
| 121 | /** attributes name for the new dataset used to build the model */ |
---|
| 122 | protected Instances m_Attributes; |
---|
| 123 | |
---|
| 124 | /** Number of iterations */ |
---|
| 125 | private int m_NumIterations = 100; |
---|
| 126 | |
---|
| 127 | /** Voting weights of models */ |
---|
| 128 | protected double[] m_Beta; |
---|
| 129 | |
---|
| 130 | /** the maximum number of boost iterations */ |
---|
| 131 | protected int m_MaxIterations = 10; |
---|
| 132 | |
---|
| 133 | /** the number of discretization bins */ |
---|
| 134 | protected int m_DiscretizeBin = 0; |
---|
| 135 | |
---|
| 136 | /** filter used for discretization */ |
---|
| 137 | protected Discretize m_Filter = null; |
---|
| 138 | |
---|
| 139 | /** filter used to convert the MI dataset into single-instance dataset */ |
---|
| 140 | protected MultiInstanceToPropositional m_ConvertToSI = new MultiInstanceToPropositional(); |
---|
| 141 | |
---|
| 142 | /** |
---|
| 143 | * Returns a string describing this filter |
---|
| 144 | * |
---|
| 145 | * @return a description of the filter suitable for |
---|
| 146 | * displaying in the explorer/experimenter gui |
---|
| 147 | */ |
---|
| 148 | public String globalInfo() { |
---|
| 149 | return |
---|
| 150 | "MI AdaBoost method, considers the geometric mean of posterior " |
---|
| 151 | + "of instances inside a bag (arithmatic mean of log-posterior) and " |
---|
| 152 | + "the expectation for a bag is taken inside the loss function.\n\n" |
---|
| 153 | + "For more information about Adaboost, see:\n\n" |
---|
| 154 | + getTechnicalInformation().toString(); |
---|
| 155 | } |
---|
| 156 | |
---|
| 157 | /** |
---|
| 158 | * Returns an instance of a TechnicalInformation object, containing |
---|
| 159 | * detailed information about the technical background of this class, |
---|
| 160 | * e.g., paper reference or book this class is based on. |
---|
| 161 | * |
---|
| 162 | * @return the technical information about this class |
---|
| 163 | */ |
---|
| 164 | public TechnicalInformation getTechnicalInformation() { |
---|
| 165 | TechnicalInformation result; |
---|
| 166 | |
---|
| 167 | result = new TechnicalInformation(Type.INPROCEEDINGS); |
---|
| 168 | result.setValue(Field.AUTHOR, "Yoav Freund and Robert E. Schapire"); |
---|
| 169 | result.setValue(Field.TITLE, "Experiments with a new boosting algorithm"); |
---|
| 170 | result.setValue(Field.BOOKTITLE, "Thirteenth International Conference on Machine Learning"); |
---|
| 171 | result.setValue(Field.YEAR, "1996"); |
---|
| 172 | result.setValue(Field.PAGES, "148-156"); |
---|
| 173 | result.setValue(Field.PUBLISHER, "Morgan Kaufmann"); |
---|
| 174 | result.setValue(Field.ADDRESS, "San Francisco"); |
---|
| 175 | |
---|
| 176 | return result; |
---|
| 177 | } |
---|
| 178 | |
---|
| 179 | /** |
---|
| 180 | * Returns an enumeration describing the available options |
---|
| 181 | * |
---|
| 182 | * @return an enumeration of all the available options |
---|
| 183 | */ |
---|
| 184 | public Enumeration listOptions() { |
---|
| 185 | Vector result = new Vector(); |
---|
| 186 | |
---|
| 187 | result.addElement(new Option( |
---|
| 188 | "\tTurn on debugging output.", |
---|
| 189 | "D", 0, "-D")); |
---|
| 190 | |
---|
| 191 | result.addElement(new Option( |
---|
| 192 | "\tThe number of bins in discretization\n" |
---|
| 193 | + "\t(default 0, no discretization)", |
---|
| 194 | "B", 1, "-B <num>")); |
---|
| 195 | |
---|
| 196 | result.addElement(new Option( |
---|
| 197 | "\tMaximum number of boost iterations.\n" |
---|
| 198 | + "\t(default 10)", |
---|
| 199 | "R", 1, "-R <num>")); |
---|
| 200 | |
---|
| 201 | result.addElement(new Option( |
---|
| 202 | "\tFull name of classifier to boost.\n" |
---|
| 203 | + "\teg: weka.classifiers.bayes.NaiveBayes", |
---|
| 204 | "W", 1, "-W <class name>")); |
---|
| 205 | |
---|
| 206 | Enumeration enu = ((OptionHandler)m_Classifier).listOptions(); |
---|
| 207 | while (enu.hasMoreElements()) { |
---|
| 208 | result.addElement(enu.nextElement()); |
---|
| 209 | } |
---|
| 210 | |
---|
| 211 | return result.elements(); |
---|
| 212 | } |
---|
| 213 | |
---|
| 214 | /** |
---|
| 215 | * Parses a given list of options. <p/> |
---|
| 216 | * |
---|
| 217 | <!-- options-start --> |
---|
| 218 | * Valid options are: <p/> |
---|
| 219 | * |
---|
| 220 | * <pre> -D |
---|
| 221 | * Turn on debugging output.</pre> |
---|
| 222 | * |
---|
| 223 | * <pre> -B <num> |
---|
| 224 | * The number of bins in discretization |
---|
| 225 | * (default 0, no discretization)</pre> |
---|
| 226 | * |
---|
| 227 | * <pre> -R <num> |
---|
| 228 | * Maximum number of boost iterations. |
---|
| 229 | * (default 10)</pre> |
---|
| 230 | * |
---|
| 231 | * <pre> -W <class name> |
---|
| 232 | * Full name of classifier to boost. |
---|
| 233 | * eg: weka.classifiers.bayes.NaiveBayes</pre> |
---|
| 234 | * |
---|
| 235 | * <pre> -D |
---|
| 236 | * If set, classifier is run in debug mode and |
---|
| 237 | * may output additional info to the console</pre> |
---|
| 238 | * |
---|
| 239 | <!-- options-end --> |
---|
| 240 | * |
---|
| 241 | * @param options the list of options as an array of strings |
---|
| 242 | * @throws Exception if an option is not supported |
---|
| 243 | */ |
---|
| 244 | public void setOptions(String[] options) throws Exception { |
---|
| 245 | setDebug(Utils.getFlag('D', options)); |
---|
| 246 | |
---|
| 247 | String bin = Utils.getOption('B', options); |
---|
| 248 | if (bin.length() != 0) { |
---|
| 249 | setDiscretizeBin(Integer.parseInt(bin)); |
---|
| 250 | } else { |
---|
| 251 | setDiscretizeBin(0); |
---|
| 252 | } |
---|
| 253 | |
---|
| 254 | String boostIterations = Utils.getOption('R', options); |
---|
| 255 | if (boostIterations.length() != 0) { |
---|
| 256 | setMaxIterations(Integer.parseInt(boostIterations)); |
---|
| 257 | } else { |
---|
| 258 | setMaxIterations(10); |
---|
| 259 | } |
---|
| 260 | |
---|
| 261 | super.setOptions(options); |
---|
| 262 | } |
---|
| 263 | |
---|
| 264 | /** |
---|
| 265 | * Gets the current settings of the classifier. |
---|
| 266 | * |
---|
| 267 | * @return an array of strings suitable for passing to setOptions |
---|
| 268 | */ |
---|
| 269 | public String[] getOptions() { |
---|
| 270 | Vector result; |
---|
| 271 | String[] options; |
---|
| 272 | int i; |
---|
| 273 | |
---|
| 274 | result = new Vector(); |
---|
| 275 | |
---|
| 276 | result.add("-R"); |
---|
| 277 | result.add("" + getMaxIterations()); |
---|
| 278 | |
---|
| 279 | result.add("-B"); |
---|
| 280 | result.add("" + getDiscretizeBin()); |
---|
| 281 | |
---|
| 282 | options = super.getOptions(); |
---|
| 283 | for (i = 0; i < options.length; i++) |
---|
| 284 | result.add(options[i]); |
---|
| 285 | |
---|
| 286 | return (String[]) result.toArray(new String[result.size()]); |
---|
| 287 | } |
---|
| 288 | |
---|
| 289 | /** |
---|
| 290 | * Returns the tip text for this property |
---|
| 291 | * |
---|
| 292 | * @return tip text for this property suitable for |
---|
| 293 | * displaying in the explorer/experimenter gui |
---|
| 294 | */ |
---|
| 295 | public String maxIterationsTipText() { |
---|
| 296 | return "The maximum number of boost iterations."; |
---|
| 297 | } |
---|
| 298 | |
---|
| 299 | /** |
---|
| 300 | * Set the maximum number of boost iterations |
---|
| 301 | * |
---|
| 302 | * @param maxIterations the maximum number of boost iterations |
---|
| 303 | */ |
---|
| 304 | public void setMaxIterations(int maxIterations) { |
---|
| 305 | m_MaxIterations = maxIterations; |
---|
| 306 | } |
---|
| 307 | |
---|
| 308 | /** |
---|
| 309 | * Get the maximum number of boost iterations |
---|
| 310 | * |
---|
| 311 | * @return the maximum number of boost iterations |
---|
| 312 | */ |
---|
| 313 | public int getMaxIterations() { |
---|
| 314 | |
---|
| 315 | return m_MaxIterations; |
---|
| 316 | } |
---|
| 317 | |
---|
| 318 | /** |
---|
| 319 | * Returns the tip text for this property |
---|
| 320 | * |
---|
| 321 | * @return tip text for this property suitable for |
---|
| 322 | * displaying in the explorer/experimenter gui |
---|
| 323 | */ |
---|
| 324 | public String discretizeBinTipText() { |
---|
| 325 | return "The number of bins in discretization."; |
---|
| 326 | } |
---|
| 327 | |
---|
| 328 | /** |
---|
| 329 | * Set the number of bins in discretization |
---|
| 330 | * |
---|
| 331 | * @param bin the number of bins in discretization |
---|
| 332 | */ |
---|
| 333 | public void setDiscretizeBin(int bin) { |
---|
| 334 | m_DiscretizeBin = bin; |
---|
| 335 | } |
---|
| 336 | |
---|
| 337 | /** |
---|
| 338 | * Get the number of bins in discretization |
---|
| 339 | * |
---|
| 340 | * @return the number of bins in discretization |
---|
| 341 | */ |
---|
| 342 | public int getDiscretizeBin() { |
---|
| 343 | return m_DiscretizeBin; |
---|
| 344 | } |
---|
| 345 | |
---|
| 346 | private class OptEng |
---|
| 347 | extends Optimization { |
---|
| 348 | |
---|
| 349 | private double[] weights, errs; |
---|
| 350 | |
---|
| 351 | public void setWeights(double[] w){ |
---|
| 352 | weights = w; |
---|
| 353 | } |
---|
| 354 | |
---|
| 355 | public void setErrs(double[] e){ |
---|
| 356 | errs = e; |
---|
| 357 | } |
---|
| 358 | |
---|
| 359 | /** |
---|
| 360 | * Evaluate objective function |
---|
| 361 | * @param x the current values of variables |
---|
| 362 | * @return the value of the objective function |
---|
| 363 | * @throws Exception if result is NaN |
---|
| 364 | */ |
---|
| 365 | protected double objectiveFunction(double[] x) throws Exception{ |
---|
| 366 | double obj=0; |
---|
| 367 | for(int i=0; i<weights.length; i++){ |
---|
| 368 | obj += weights[i]*Math.exp(x[0]*(2.0*errs[i]-1.0)); |
---|
| 369 | if(Double.isNaN(obj)) |
---|
| 370 | throw new Exception("Objective function value is NaN!"); |
---|
| 371 | |
---|
| 372 | } |
---|
| 373 | return obj; |
---|
| 374 | } |
---|
| 375 | |
---|
| 376 | /** |
---|
| 377 | * Evaluate Jacobian vector |
---|
| 378 | * @param x the current values of variables |
---|
| 379 | * @return the gradient vector |
---|
| 380 | * @throws Exception if gradient is NaN |
---|
| 381 | */ |
---|
| 382 | protected double[] evaluateGradient(double[] x) throws Exception{ |
---|
| 383 | double[] grad = new double[1]; |
---|
| 384 | for(int i=0; i<weights.length; i++){ |
---|
| 385 | grad[0] += weights[i]*(2.0*errs[i]-1.0)*Math.exp(x[0]*(2.0*errs[i]-1.0)); |
---|
| 386 | if(Double.isNaN(grad[0])) |
---|
| 387 | throw new Exception("Gradient is NaN!"); |
---|
| 388 | |
---|
| 389 | } |
---|
| 390 | return grad; |
---|
| 391 | } |
---|
| 392 | |
---|
| 393 | /** |
---|
| 394 | * Returns the revision string. |
---|
| 395 | * |
---|
| 396 | * @return the revision |
---|
| 397 | */ |
---|
| 398 | public String getRevision() { |
---|
| 399 | return RevisionUtils.extract("$Revision: 5928 $"); |
---|
| 400 | } |
---|
| 401 | } |
---|
| 402 | |
---|
| 403 | /** |
---|
| 404 | * Returns default capabilities of the classifier. |
---|
| 405 | * |
---|
| 406 | * @return the capabilities of this classifier |
---|
| 407 | */ |
---|
| 408 | public Capabilities getCapabilities() { |
---|
| 409 | Capabilities result = super.getCapabilities(); |
---|
| 410 | |
---|
| 411 | // attributes |
---|
| 412 | result.enable(Capability.NOMINAL_ATTRIBUTES); |
---|
| 413 | result.enable(Capability.RELATIONAL_ATTRIBUTES); |
---|
| 414 | result.enable(Capability.MISSING_VALUES); |
---|
| 415 | |
---|
| 416 | // class |
---|
| 417 | result.disableAllClasses(); |
---|
| 418 | result.disableAllClassDependencies(); |
---|
| 419 | if (super.getCapabilities().handles(Capability.BINARY_CLASS)) |
---|
| 420 | result.enable(Capability.BINARY_CLASS); |
---|
| 421 | result.enable(Capability.MISSING_CLASS_VALUES); |
---|
| 422 | |
---|
| 423 | // other |
---|
| 424 | result.enable(Capability.ONLY_MULTIINSTANCE); |
---|
| 425 | |
---|
| 426 | return result; |
---|
| 427 | } |
---|
| 428 | |
---|
| 429 | /** |
---|
| 430 | * Returns the capabilities of this multi-instance classifier for the |
---|
| 431 | * relational data. |
---|
| 432 | * |
---|
| 433 | * @return the capabilities of this object |
---|
| 434 | * @see Capabilities |
---|
| 435 | */ |
---|
| 436 | public Capabilities getMultiInstanceCapabilities() { |
---|
| 437 | Capabilities result = super.getCapabilities(); |
---|
| 438 | |
---|
| 439 | // class |
---|
| 440 | result.disableAllClasses(); |
---|
| 441 | result.enable(Capability.NO_CLASS); |
---|
| 442 | |
---|
| 443 | return result; |
---|
| 444 | } |
---|
| 445 | |
---|
| 446 | /** |
---|
| 447 | * Builds the classifier |
---|
| 448 | * |
---|
| 449 | * @param exps the training data to be used for generating the |
---|
| 450 | * boosted classifier. |
---|
| 451 | * @throws Exception if the classifier could not be built successfully |
---|
| 452 | */ |
---|
| 453 | public void buildClassifier(Instances exps) throws Exception { |
---|
| 454 | |
---|
| 455 | // can classifier handle the data? |
---|
| 456 | getCapabilities().testWithFail(exps); |
---|
| 457 | |
---|
| 458 | // remove instances with missing class |
---|
| 459 | Instances train = new Instances(exps); |
---|
| 460 | train.deleteWithMissingClass(); |
---|
| 461 | |
---|
| 462 | m_NumClasses = train.numClasses(); |
---|
| 463 | m_NumIterations = m_MaxIterations; |
---|
| 464 | |
---|
| 465 | if (m_Classifier == null) |
---|
| 466 | throw new Exception("A base classifier has not been specified!"); |
---|
| 467 | if(!(m_Classifier instanceof WeightedInstancesHandler)) |
---|
| 468 | throw new Exception("Base classifier cannot handle weighted instances!"); |
---|
| 469 | |
---|
| 470 | m_Models = AbstractClassifier.makeCopies(m_Classifier, getMaxIterations()); |
---|
| 471 | if(m_Debug) |
---|
| 472 | System.err.println("Base classifier: "+m_Classifier.getClass().getName()); |
---|
| 473 | |
---|
| 474 | m_Beta = new double[m_NumIterations]; |
---|
| 475 | |
---|
| 476 | /* modified by Lin Dong. (use MIToSingleInstance filter to convert the MI datasets) */ |
---|
| 477 | |
---|
| 478 | //Initialize the bags' weights |
---|
| 479 | double N = (double)train.numInstances(), sumNi=0; |
---|
| 480 | for(int i=0; i<N; i++) |
---|
| 481 | sumNi += train.instance(i).relationalValue(1).numInstances(); |
---|
| 482 | for(int i=0; i<N; i++){ |
---|
| 483 | train.instance(i).setWeight(sumNi/N); |
---|
| 484 | } |
---|
| 485 | |
---|
| 486 | //convert the training dataset into single-instance dataset |
---|
| 487 | m_ConvertToSI.setInputFormat(train); |
---|
| 488 | Instances data = Filter.useFilter( train, m_ConvertToSI); |
---|
| 489 | data.deleteAttributeAt(0); //remove the bagIndex attribute; |
---|
| 490 | |
---|
| 491 | |
---|
| 492 | // Assume the order of the instances are preserved in the Discretize filter |
---|
| 493 | if(m_DiscretizeBin > 0){ |
---|
| 494 | m_Filter = new Discretize(); |
---|
| 495 | m_Filter.setInputFormat(new Instances(data, 0)); |
---|
| 496 | m_Filter.setBins(m_DiscretizeBin); |
---|
| 497 | data = Filter.useFilter(data, m_Filter); |
---|
| 498 | } |
---|
| 499 | |
---|
| 500 | // Main algorithm |
---|
| 501 | int dataIdx; |
---|
| 502 | iterations: |
---|
| 503 | for(int m=0; m < m_MaxIterations; m++){ |
---|
| 504 | if(m_Debug) |
---|
| 505 | System.err.println("\nIteration "+m); |
---|
| 506 | |
---|
| 507 | |
---|
| 508 | // Build a model |
---|
| 509 | m_Models[m].buildClassifier(data); |
---|
| 510 | |
---|
| 511 | // Prediction of each bag |
---|
| 512 | double[] err=new double[(int)N], weights=new double[(int)N]; |
---|
| 513 | boolean perfect = true, tooWrong=true; |
---|
| 514 | dataIdx = 0; |
---|
| 515 | for(int n=0; n<N; n++){ |
---|
| 516 | Instance exn = train.instance(n); |
---|
| 517 | // Prediction of each instance and the predicted class distribution |
---|
| 518 | // of the bag |
---|
| 519 | double nn = (double)exn.relationalValue(1).numInstances(); |
---|
| 520 | for(int p=0; p<nn; p++){ |
---|
| 521 | Instance testIns = data.instance(dataIdx++); |
---|
| 522 | if((int)m_Models[m].classifyInstance(testIns) |
---|
| 523 | != (int)exn.classValue()) // Weighted instance-wise 0-1 errors |
---|
| 524 | err[n] ++; |
---|
| 525 | } |
---|
| 526 | weights[n] = exn.weight(); |
---|
| 527 | err[n] /= nn; |
---|
| 528 | if(err[n] > 0.5) |
---|
| 529 | perfect = false; |
---|
| 530 | if(err[n] < 0.5) |
---|
| 531 | tooWrong = false; |
---|
| 532 | } |
---|
| 533 | |
---|
| 534 | if(perfect || tooWrong){ // No or 100% classification error, cannot find beta |
---|
| 535 | if (m == 0) |
---|
| 536 | m_Beta[m] = 1.0; |
---|
| 537 | else |
---|
| 538 | m_Beta[m] = 0; |
---|
| 539 | m_NumIterations = m+1; |
---|
| 540 | if(m_Debug) System.err.println("No errors"); |
---|
| 541 | break iterations; |
---|
| 542 | } |
---|
| 543 | |
---|
| 544 | double[] x = new double[1]; |
---|
| 545 | x[0] = 0; |
---|
| 546 | double[][] b = new double[2][x.length]; |
---|
| 547 | b[0][0] = Double.NaN; |
---|
| 548 | b[1][0] = Double.NaN; |
---|
| 549 | |
---|
| 550 | OptEng opt = new OptEng(); |
---|
| 551 | opt.setWeights(weights); |
---|
| 552 | opt.setErrs(err); |
---|
| 553 | //opt.setDebug(m_Debug); |
---|
| 554 | if (m_Debug) |
---|
| 555 | System.out.println("Start searching for c... "); |
---|
| 556 | x = opt.findArgmin(x, b); |
---|
| 557 | while(x==null){ |
---|
| 558 | x = opt.getVarbValues(); |
---|
| 559 | if (m_Debug) |
---|
| 560 | System.out.println("200 iterations finished, not enough!"); |
---|
| 561 | x = opt.findArgmin(x, b); |
---|
| 562 | } |
---|
| 563 | if (m_Debug) |
---|
| 564 | System.out.println("Finished."); |
---|
| 565 | m_Beta[m] = x[0]; |
---|
| 566 | |
---|
| 567 | if(m_Debug) |
---|
| 568 | System.err.println("c = "+m_Beta[m]); |
---|
| 569 | |
---|
| 570 | // Stop if error too small or error too big and ignore this model |
---|
| 571 | if (Double.isInfinite(m_Beta[m]) |
---|
| 572 | || Utils.smOrEq(m_Beta[m], 0) |
---|
| 573 | ) { |
---|
| 574 | if (m == 0) |
---|
| 575 | m_Beta[m] = 1.0; |
---|
| 576 | else |
---|
| 577 | m_Beta[m] = 0; |
---|
| 578 | m_NumIterations = m+1; |
---|
| 579 | if(m_Debug) |
---|
| 580 | System.err.println("Errors out of range!"); |
---|
| 581 | break iterations; |
---|
| 582 | } |
---|
| 583 | |
---|
| 584 | // Update weights of data and class label of wfData |
---|
| 585 | dataIdx=0; |
---|
| 586 | double totWeights=0; |
---|
| 587 | for(int r=0; r<N; r++){ |
---|
| 588 | Instance exr = train.instance(r); |
---|
| 589 | exr.setWeight(weights[r]*Math.exp(m_Beta[m]*(2.0*err[r]-1.0))); |
---|
| 590 | totWeights += exr.weight(); |
---|
| 591 | } |
---|
| 592 | |
---|
| 593 | if(m_Debug) |
---|
| 594 | System.err.println("Total weights = "+totWeights); |
---|
| 595 | |
---|
| 596 | for(int r=0; r<N; r++){ |
---|
| 597 | Instance exr = train.instance(r); |
---|
| 598 | double num = (double)exr.relationalValue(1).numInstances(); |
---|
| 599 | exr.setWeight(sumNi*exr.weight()/totWeights); |
---|
| 600 | //if(m_Debug) |
---|
| 601 | // System.err.print("\nExemplar "+r+"="+exr.weight()+": \t"); |
---|
| 602 | for(int s=0; s<num; s++){ |
---|
| 603 | Instance inss = data.instance(dataIdx); |
---|
| 604 | inss.setWeight(exr.weight()/num); |
---|
| 605 | // if(m_Debug) |
---|
| 606 | // System.err.print("instance "+s+"="+inss.weight()+ |
---|
| 607 | // "|ew*iw*sumNi="+data.instance(dataIdx).weight()+"\t"); |
---|
| 608 | if(Double.isNaN(inss.weight())) |
---|
| 609 | throw new Exception("instance "+s+" in bag "+r+" has weight NaN!"); |
---|
| 610 | dataIdx++; |
---|
| 611 | } |
---|
| 612 | //if(m_Debug) |
---|
| 613 | // System.err.println(); |
---|
| 614 | } |
---|
| 615 | } |
---|
| 616 | } |
---|
| 617 | |
---|
| 618 | /** |
---|
| 619 | * Computes the distribution for a given exemplar |
---|
| 620 | * |
---|
| 621 | * @param exmp the exemplar for which distribution is computed |
---|
| 622 | * @return the classification |
---|
| 623 | * @throws Exception if the distribution can't be computed successfully |
---|
| 624 | */ |
---|
| 625 | public double[] distributionForInstance(Instance exmp) |
---|
| 626 | throws Exception { |
---|
| 627 | |
---|
| 628 | double[] rt = new double[m_NumClasses]; |
---|
| 629 | |
---|
| 630 | Instances insts = new Instances(exmp.dataset(), 0); |
---|
| 631 | insts.add(exmp); |
---|
| 632 | |
---|
| 633 | // convert the training dataset into single-instance dataset |
---|
| 634 | insts = Filter.useFilter( insts, m_ConvertToSI); |
---|
| 635 | insts.deleteAttributeAt(0); //remove the bagIndex attribute |
---|
| 636 | |
---|
| 637 | double n = insts.numInstances(); |
---|
| 638 | |
---|
| 639 | if(m_DiscretizeBin > 0) |
---|
| 640 | insts = Filter.useFilter(insts, m_Filter); |
---|
| 641 | |
---|
| 642 | for(int y=0; y<n; y++){ |
---|
| 643 | Instance ins = insts.instance(y); |
---|
| 644 | for(int x=0; x<m_NumIterations; x++){ |
---|
| 645 | rt[(int)m_Models[x].classifyInstance(ins)] += m_Beta[x]/n; |
---|
| 646 | } |
---|
| 647 | } |
---|
| 648 | |
---|
| 649 | for(int i=0; i<rt.length; i++) |
---|
| 650 | rt[i] = Math.exp(rt[i]); |
---|
| 651 | |
---|
| 652 | Utils.normalize(rt); |
---|
| 653 | return rt; |
---|
| 654 | } |
---|
| 655 | |
---|
| 656 | /** |
---|
| 657 | * Gets a string describing the classifier. |
---|
| 658 | * |
---|
| 659 | * @return a string describing the classifer built. |
---|
| 660 | */ |
---|
| 661 | public String toString() { |
---|
| 662 | |
---|
| 663 | if (m_Models == null) { |
---|
| 664 | return "No model built yet!"; |
---|
| 665 | } |
---|
| 666 | StringBuffer text = new StringBuffer(); |
---|
| 667 | text.append("MIBoost: number of bins in discretization = "+m_DiscretizeBin+"\n"); |
---|
| 668 | if (m_NumIterations == 0) { |
---|
| 669 | text.append("No model built yet.\n"); |
---|
| 670 | } else if (m_NumIterations == 1) { |
---|
| 671 | text.append("No boosting possible, one classifier used: Weight = " |
---|
| 672 | + Utils.roundDouble(m_Beta[0], 2)+"\n"); |
---|
| 673 | text.append("Base classifiers:\n"+m_Models[0].toString()); |
---|
| 674 | } else { |
---|
| 675 | text.append("Base classifiers and their weights: \n"); |
---|
| 676 | for (int i = 0; i < m_NumIterations ; i++) { |
---|
| 677 | text.append("\n\n"+i+": Weight = " + Utils.roundDouble(m_Beta[i], 2) |
---|
| 678 | +"\nBase classifier:\n"+m_Models[i].toString() ); |
---|
| 679 | } |
---|
| 680 | } |
---|
| 681 | |
---|
| 682 | text.append("\n\nNumber of performed Iterations: " |
---|
| 683 | + m_NumIterations + "\n"); |
---|
| 684 | |
---|
| 685 | return text.toString(); |
---|
| 686 | } |
---|
| 687 | |
---|
| 688 | /** |
---|
| 689 | * Returns the revision string. |
---|
| 690 | * |
---|
| 691 | * @return the revision |
---|
| 692 | */ |
---|
| 693 | public String getRevision() { |
---|
| 694 | return RevisionUtils.extract("$Revision: 5928 $"); |
---|
| 695 | } |
---|
| 696 | |
---|
| 697 | /** |
---|
| 698 | * Main method for testing this class. |
---|
| 699 | * |
---|
| 700 | * @param argv should contain the command line arguments to the |
---|
| 701 | * scheme (see Evaluation) |
---|
| 702 | */ |
---|
| 703 | public static void main(String[] argv) { |
---|
| 704 | runClassifier(new MIBoost(), argv); |
---|
| 705 | } |
---|
| 706 | } |
---|