| 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 | * HillClimber.java |
|---|
| 19 | * Copyright (C) 2004 University of Waikato, Hamilton, New Zealand |
|---|
| 20 | * |
|---|
| 21 | */ |
|---|
| 22 | |
|---|
| 23 | package weka.classifiers.bayes.net.search.global; |
|---|
| 24 | |
|---|
| 25 | import weka.classifiers.bayes.BayesNet; |
|---|
| 26 | import weka.classifiers.bayes.net.ParentSet; |
|---|
| 27 | import weka.core.Instances; |
|---|
| 28 | import weka.core.Option; |
|---|
| 29 | import weka.core.RevisionHandler; |
|---|
| 30 | import weka.core.RevisionUtils; |
|---|
| 31 | import weka.core.Utils; |
|---|
| 32 | |
|---|
| 33 | import java.io.Serializable; |
|---|
| 34 | import java.util.Enumeration; |
|---|
| 35 | import java.util.Vector; |
|---|
| 36 | |
|---|
| 37 | /** |
|---|
| 38 | <!-- globalinfo-start --> |
|---|
| 39 | * This Bayes Network learning algorithm uses a hill climbing algorithm adding, deleting and reversing arcs. The search is not restricted by an order on the variables (unlike K2). The difference with B and B2 is that this hill climber also considers arrows part of the naive Bayes structure for deletion. |
|---|
| 40 | * <p/> |
|---|
| 41 | <!-- globalinfo-end --> |
|---|
| 42 | * |
|---|
| 43 | <!-- options-start --> |
|---|
| 44 | * Valid options are: <p/> |
|---|
| 45 | * |
|---|
| 46 | * <pre> -P <nr of parents> |
|---|
| 47 | * Maximum number of parents</pre> |
|---|
| 48 | * |
|---|
| 49 | * <pre> -R |
|---|
| 50 | * Use arc reversal operation. |
|---|
| 51 | * (default false)</pre> |
|---|
| 52 | * |
|---|
| 53 | * <pre> -N |
|---|
| 54 | * Initial structure is empty (instead of Naive Bayes)</pre> |
|---|
| 55 | * |
|---|
| 56 | * <pre> -mbc |
|---|
| 57 | * Applies a Markov Blanket correction to the network structure, |
|---|
| 58 | * after a network structure is learned. This ensures that all |
|---|
| 59 | * nodes in the network are part of the Markov blanket of the |
|---|
| 60 | * classifier node.</pre> |
|---|
| 61 | * |
|---|
| 62 | * <pre> -S [LOO-CV|k-Fold-CV|Cumulative-CV] |
|---|
| 63 | * Score type (LOO-CV,k-Fold-CV,Cumulative-CV)</pre> |
|---|
| 64 | * |
|---|
| 65 | * <pre> -Q |
|---|
| 66 | * Use probabilistic or 0/1 scoring. |
|---|
| 67 | * (default probabilistic scoring)</pre> |
|---|
| 68 | * |
|---|
| 69 | <!-- options-end --> |
|---|
| 70 | * |
|---|
| 71 | * @author Remco Bouckaert (rrb@xm.co.nz) |
|---|
| 72 | * @version $Revision: 1.9 $ |
|---|
| 73 | */ |
|---|
| 74 | public class HillClimber |
|---|
| 75 | extends GlobalScoreSearchAlgorithm { |
|---|
| 76 | |
|---|
| 77 | /** for serialization */ |
|---|
| 78 | static final long serialVersionUID = -3885042888195820149L; |
|---|
| 79 | |
|---|
| 80 | /** |
|---|
| 81 | * the Operation class contains info on operations performed |
|---|
| 82 | * on the current Bayesian network. |
|---|
| 83 | */ |
|---|
| 84 | class Operation |
|---|
| 85 | implements Serializable, RevisionHandler { |
|---|
| 86 | |
|---|
| 87 | /** for serialization */ |
|---|
| 88 | static final long serialVersionUID = -2934970456587374967L; |
|---|
| 89 | |
|---|
| 90 | // constants indicating the type of an operation |
|---|
| 91 | final static int OPERATION_ADD = 0; |
|---|
| 92 | final static int OPERATION_DEL = 1; |
|---|
| 93 | final static int OPERATION_REVERSE = 2; |
|---|
| 94 | |
|---|
| 95 | /** c'tor **/ |
|---|
| 96 | public Operation() { |
|---|
| 97 | } |
|---|
| 98 | |
|---|
| 99 | /** c'tor + initializers |
|---|
| 100 | * |
|---|
| 101 | * @param nTail |
|---|
| 102 | * @param nHead |
|---|
| 103 | * @param nOperation |
|---|
| 104 | */ |
|---|
| 105 | public Operation(int nTail, int nHead, int nOperation) { |
|---|
| 106 | m_nHead = nHead; |
|---|
| 107 | m_nTail = nTail; |
|---|
| 108 | m_nOperation = nOperation; |
|---|
| 109 | } |
|---|
| 110 | /** compare this operation with another |
|---|
| 111 | * @param other operation to compare with |
|---|
| 112 | * @return true if operation is the same |
|---|
| 113 | */ |
|---|
| 114 | public boolean equals(Operation other) { |
|---|
| 115 | if (other == null) { |
|---|
| 116 | return false; |
|---|
| 117 | } |
|---|
| 118 | return (( m_nOperation == other.m_nOperation) && |
|---|
| 119 | (m_nHead == other.m_nHead) && |
|---|
| 120 | (m_nTail == other.m_nTail)); |
|---|
| 121 | } // equals |
|---|
| 122 | /** number of the tail node **/ |
|---|
| 123 | public int m_nTail; |
|---|
| 124 | /** number of the head node **/ |
|---|
| 125 | public int m_nHead; |
|---|
| 126 | /** type of operation (ADD, DEL, REVERSE) **/ |
|---|
| 127 | public int m_nOperation; |
|---|
| 128 | /** change of score due to this operation **/ |
|---|
| 129 | public double m_fScore = -1E100; |
|---|
| 130 | |
|---|
| 131 | /** |
|---|
| 132 | * Returns the revision string. |
|---|
| 133 | * |
|---|
| 134 | * @return the revision |
|---|
| 135 | */ |
|---|
| 136 | public String getRevision() { |
|---|
| 137 | return RevisionUtils.extract("$Revision: 1.9 $"); |
|---|
| 138 | } |
|---|
| 139 | } // class Operation |
|---|
| 140 | |
|---|
| 141 | /** use the arc reversal operator **/ |
|---|
| 142 | boolean m_bUseArcReversal = false; |
|---|
| 143 | |
|---|
| 144 | /** |
|---|
| 145 | * search determines the network structure/graph of the network |
|---|
| 146 | * with the Taby algorithm. |
|---|
| 147 | * |
|---|
| 148 | * @param bayesNet the network to search |
|---|
| 149 | * @param instances the instances to work with |
|---|
| 150 | * @throws Exception if something goes wrong |
|---|
| 151 | */ |
|---|
| 152 | protected void search(BayesNet bayesNet, Instances instances) throws Exception { |
|---|
| 153 | m_BayesNet = bayesNet; |
|---|
| 154 | double fScore = calcScore(bayesNet); |
|---|
| 155 | // go do the search |
|---|
| 156 | Operation oOperation = getOptimalOperation(bayesNet, instances); |
|---|
| 157 | while ((oOperation != null) && (oOperation.m_fScore > fScore)) { |
|---|
| 158 | performOperation(bayesNet, instances, oOperation); |
|---|
| 159 | fScore = oOperation.m_fScore; |
|---|
| 160 | oOperation = getOptimalOperation(bayesNet, instances); |
|---|
| 161 | } |
|---|
| 162 | } // search |
|---|
| 163 | |
|---|
| 164 | |
|---|
| 165 | |
|---|
| 166 | /** check whether the operation is not in the forbidden. |
|---|
| 167 | * For base hill climber, there are no restrictions on operations, |
|---|
| 168 | * so we always return true. |
|---|
| 169 | * @param oOperation operation to be checked |
|---|
| 170 | * @return true if operation is not in the tabu list |
|---|
| 171 | */ |
|---|
| 172 | boolean isNotTabu(Operation oOperation) { |
|---|
| 173 | return true; |
|---|
| 174 | } // isNotTabu |
|---|
| 175 | |
|---|
| 176 | /** |
|---|
| 177 | * getOptimalOperation finds the optimal operation that can be performed |
|---|
| 178 | * on the Bayes network that is not in the tabu list. |
|---|
| 179 | * |
|---|
| 180 | * @param bayesNet Bayes network to apply operation on |
|---|
| 181 | * @param instances data set to learn from |
|---|
| 182 | * @return optimal operation found |
|---|
| 183 | * @throws Exception if something goes wrong |
|---|
| 184 | */ |
|---|
| 185 | Operation getOptimalOperation(BayesNet bayesNet, Instances instances) throws Exception { |
|---|
| 186 | Operation oBestOperation = new Operation(); |
|---|
| 187 | |
|---|
| 188 | // Add??? |
|---|
| 189 | oBestOperation = findBestArcToAdd(bayesNet, instances, oBestOperation); |
|---|
| 190 | // Delete??? |
|---|
| 191 | oBestOperation = findBestArcToDelete(bayesNet, instances, oBestOperation); |
|---|
| 192 | // Reverse??? |
|---|
| 193 | if (getUseArcReversal()) { |
|---|
| 194 | oBestOperation = findBestArcToReverse(bayesNet, instances, oBestOperation); |
|---|
| 195 | } |
|---|
| 196 | |
|---|
| 197 | // did we find something? |
|---|
| 198 | if (oBestOperation.m_fScore == -1E100) { |
|---|
| 199 | return null; |
|---|
| 200 | } |
|---|
| 201 | |
|---|
| 202 | return oBestOperation; |
|---|
| 203 | } // getOptimalOperation |
|---|
| 204 | |
|---|
| 205 | /** performOperation applies an operation |
|---|
| 206 | * on the Bayes network and update the cache. |
|---|
| 207 | * |
|---|
| 208 | * @param bayesNet Bayes network to apply operation on |
|---|
| 209 | * @param instances data set to learn from |
|---|
| 210 | * @param oOperation operation to perform |
|---|
| 211 | * @throws Exception if something goes wrong |
|---|
| 212 | */ |
|---|
| 213 | void performOperation(BayesNet bayesNet, Instances instances, Operation oOperation) throws Exception { |
|---|
| 214 | // perform operation |
|---|
| 215 | switch (oOperation.m_nOperation) { |
|---|
| 216 | case Operation.OPERATION_ADD: |
|---|
| 217 | applyArcAddition(bayesNet, oOperation.m_nHead, oOperation.m_nTail, instances); |
|---|
| 218 | if (bayesNet.getDebug()) { |
|---|
| 219 | System.out.print("Add " + oOperation.m_nHead + " -> " + oOperation.m_nTail); |
|---|
| 220 | } |
|---|
| 221 | break; |
|---|
| 222 | case Operation.OPERATION_DEL: |
|---|
| 223 | applyArcDeletion(bayesNet, oOperation.m_nHead, oOperation.m_nTail, instances); |
|---|
| 224 | if (bayesNet.getDebug()) { |
|---|
| 225 | System.out.print("Del " + oOperation.m_nHead + " -> " + oOperation.m_nTail); |
|---|
| 226 | } |
|---|
| 227 | break; |
|---|
| 228 | case Operation.OPERATION_REVERSE: |
|---|
| 229 | applyArcDeletion(bayesNet, oOperation.m_nHead, oOperation.m_nTail, instances); |
|---|
| 230 | applyArcAddition(bayesNet, oOperation.m_nTail, oOperation.m_nHead, instances); |
|---|
| 231 | if (bayesNet.getDebug()) { |
|---|
| 232 | System.out.print("Rev " + oOperation.m_nHead+ " -> " + oOperation.m_nTail); |
|---|
| 233 | } |
|---|
| 234 | break; |
|---|
| 235 | } |
|---|
| 236 | } // performOperation |
|---|
| 237 | |
|---|
| 238 | /** |
|---|
| 239 | * |
|---|
| 240 | * @param bayesNet |
|---|
| 241 | * @param iHead |
|---|
| 242 | * @param iTail |
|---|
| 243 | * @param instances |
|---|
| 244 | */ |
|---|
| 245 | void applyArcAddition(BayesNet bayesNet, int iHead, int iTail, Instances instances) { |
|---|
| 246 | ParentSet bestParentSet = bayesNet.getParentSet(iHead); |
|---|
| 247 | bestParentSet.addParent(iTail, instances); |
|---|
| 248 | } // applyArcAddition |
|---|
| 249 | |
|---|
| 250 | /** |
|---|
| 251 | * |
|---|
| 252 | * @param bayesNet |
|---|
| 253 | * @param iHead |
|---|
| 254 | * @param iTail |
|---|
| 255 | * @param instances |
|---|
| 256 | */ |
|---|
| 257 | void applyArcDeletion(BayesNet bayesNet, int iHead, int iTail, Instances instances) { |
|---|
| 258 | ParentSet bestParentSet = bayesNet.getParentSet(iHead); |
|---|
| 259 | bestParentSet.deleteParent(iTail, instances); |
|---|
| 260 | } // applyArcAddition |
|---|
| 261 | |
|---|
| 262 | |
|---|
| 263 | /** |
|---|
| 264 | * find best (or least bad) arc addition operation |
|---|
| 265 | * |
|---|
| 266 | * @param bayesNet Bayes network to add arc to |
|---|
| 267 | * @param instances data set |
|---|
| 268 | * @param oBestOperation |
|---|
| 269 | * @return Operation containing best arc to add, or null if no arc addition is allowed |
|---|
| 270 | * (this can happen if any arc addition introduces a cycle, or all parent sets are filled |
|---|
| 271 | * up to the maximum nr of parents). |
|---|
| 272 | * @throws Exception if something goes wrong |
|---|
| 273 | */ |
|---|
| 274 | Operation findBestArcToAdd(BayesNet bayesNet, Instances instances, Operation oBestOperation) throws Exception { |
|---|
| 275 | int nNrOfAtts = instances.numAttributes(); |
|---|
| 276 | // find best arc to add |
|---|
| 277 | for (int iAttributeHead = 0; iAttributeHead < nNrOfAtts; iAttributeHead++) { |
|---|
| 278 | if (bayesNet.getParentSet(iAttributeHead).getNrOfParents() < m_nMaxNrOfParents) { |
|---|
| 279 | for (int iAttributeTail = 0; iAttributeTail < nNrOfAtts; iAttributeTail++) { |
|---|
| 280 | if (addArcMakesSense(bayesNet, instances, iAttributeHead, iAttributeTail)) { |
|---|
| 281 | Operation oOperation = new Operation(iAttributeTail, iAttributeHead, Operation.OPERATION_ADD); |
|---|
| 282 | double fScore = calcScoreWithExtraParent(oOperation.m_nHead, oOperation.m_nTail); |
|---|
| 283 | if (fScore > oBestOperation.m_fScore) { |
|---|
| 284 | if (isNotTabu(oOperation)) { |
|---|
| 285 | oBestOperation = oOperation; |
|---|
| 286 | oBestOperation.m_fScore = fScore; |
|---|
| 287 | } |
|---|
| 288 | } |
|---|
| 289 | } |
|---|
| 290 | } |
|---|
| 291 | } |
|---|
| 292 | } |
|---|
| 293 | return oBestOperation; |
|---|
| 294 | } // findBestArcToAdd |
|---|
| 295 | |
|---|
| 296 | /** |
|---|
| 297 | * find best (or least bad) arc deletion operation |
|---|
| 298 | * |
|---|
| 299 | * @param bayesNet Bayes network to delete arc from |
|---|
| 300 | * @param instances data set |
|---|
| 301 | * @param oBestOperation |
|---|
| 302 | * @return Operation containing best arc to delete, or null if no deletion can be made |
|---|
| 303 | * (happens when there is no arc in the network yet). |
|---|
| 304 | * @throws Exception of something goes wrong |
|---|
| 305 | */ |
|---|
| 306 | Operation findBestArcToDelete(BayesNet bayesNet, Instances instances, Operation oBestOperation) throws Exception { |
|---|
| 307 | int nNrOfAtts = instances.numAttributes(); |
|---|
| 308 | // find best arc to delete |
|---|
| 309 | for (int iNode = 0; iNode < nNrOfAtts; iNode++) { |
|---|
| 310 | ParentSet parentSet = bayesNet.getParentSet(iNode); |
|---|
| 311 | for (int iParent = 0; iParent < parentSet.getNrOfParents(); iParent++) { |
|---|
| 312 | Operation oOperation = new Operation(parentSet.getParent(iParent), iNode, Operation.OPERATION_DEL); |
|---|
| 313 | double fScore = calcScoreWithMissingParent(oOperation.m_nHead, oOperation.m_nTail); |
|---|
| 314 | if (fScore > oBestOperation.m_fScore) { |
|---|
| 315 | if (isNotTabu(oOperation)) { |
|---|
| 316 | oBestOperation = oOperation; |
|---|
| 317 | oBestOperation.m_fScore = fScore; |
|---|
| 318 | } |
|---|
| 319 | } |
|---|
| 320 | } |
|---|
| 321 | } |
|---|
| 322 | return oBestOperation; |
|---|
| 323 | } // findBestArcToDelete |
|---|
| 324 | |
|---|
| 325 | /** |
|---|
| 326 | * find best (or least bad) arc reversal operation |
|---|
| 327 | * |
|---|
| 328 | * @param bayesNet Bayes network to reverse arc in |
|---|
| 329 | * @param instances data set |
|---|
| 330 | * @param oBestOperation |
|---|
| 331 | * @return Operation containing best arc to reverse, or null if no reversal is allowed |
|---|
| 332 | * (happens if there is no arc in the network yet, or when any such reversal introduces |
|---|
| 333 | * a cycle). |
|---|
| 334 | * @throws Exception if something goes wrong |
|---|
| 335 | */ |
|---|
| 336 | Operation findBestArcToReverse(BayesNet bayesNet, Instances instances, Operation oBestOperation) throws Exception { |
|---|
| 337 | int nNrOfAtts = instances.numAttributes(); |
|---|
| 338 | // find best arc to reverse |
|---|
| 339 | for (int iNode = 0; iNode < nNrOfAtts; iNode++) { |
|---|
| 340 | ParentSet parentSet = bayesNet.getParentSet(iNode); |
|---|
| 341 | for (int iParent = 0; iParent < parentSet.getNrOfParents(); iParent++) { |
|---|
| 342 | int iTail = parentSet.getParent(iParent); |
|---|
| 343 | // is reversal allowed? |
|---|
| 344 | if (reverseArcMakesSense(bayesNet, instances, iNode, iTail) && |
|---|
| 345 | bayesNet.getParentSet(iTail).getNrOfParents() < m_nMaxNrOfParents) { |
|---|
| 346 | // go check if reversal results in the best step forward |
|---|
| 347 | Operation oOperation = new Operation(parentSet.getParent(iParent), iNode, Operation.OPERATION_REVERSE); |
|---|
| 348 | double fScore = calcScoreWithReversedParent(oOperation.m_nHead, oOperation.m_nTail); |
|---|
| 349 | if (fScore > oBestOperation.m_fScore) { |
|---|
| 350 | if (isNotTabu(oOperation)) { |
|---|
| 351 | oBestOperation = oOperation; |
|---|
| 352 | oBestOperation.m_fScore = fScore; |
|---|
| 353 | } |
|---|
| 354 | } |
|---|
| 355 | } |
|---|
| 356 | } |
|---|
| 357 | } |
|---|
| 358 | return oBestOperation; |
|---|
| 359 | } // findBestArcToReverse |
|---|
| 360 | |
|---|
| 361 | |
|---|
| 362 | /** |
|---|
| 363 | * Sets the max number of parents |
|---|
| 364 | * |
|---|
| 365 | * @param nMaxNrOfParents the max number of parents |
|---|
| 366 | */ |
|---|
| 367 | public void setMaxNrOfParents(int nMaxNrOfParents) { |
|---|
| 368 | m_nMaxNrOfParents = nMaxNrOfParents; |
|---|
| 369 | } |
|---|
| 370 | |
|---|
| 371 | /** |
|---|
| 372 | * Gets the max number of parents. |
|---|
| 373 | * |
|---|
| 374 | * @return the max number of parents |
|---|
| 375 | */ |
|---|
| 376 | public int getMaxNrOfParents() { |
|---|
| 377 | return m_nMaxNrOfParents; |
|---|
| 378 | } |
|---|
| 379 | |
|---|
| 380 | /** |
|---|
| 381 | * Returns an enumeration describing the available options. |
|---|
| 382 | * |
|---|
| 383 | * @return an enumeration of all the available options. |
|---|
| 384 | */ |
|---|
| 385 | public Enumeration listOptions() { |
|---|
| 386 | Vector newVector = new Vector(2); |
|---|
| 387 | |
|---|
| 388 | newVector.addElement(new Option("\tMaximum number of parents", "P", 1, "-P <nr of parents>")); |
|---|
| 389 | newVector.addElement(new Option("\tUse arc reversal operation.\n\t(default false)", "R", 0, "-R")); |
|---|
| 390 | newVector.addElement(new Option("\tInitial structure is empty (instead of Naive Bayes)", "N", 0, "-N")); |
|---|
| 391 | |
|---|
| 392 | Enumeration enu = super.listOptions(); |
|---|
| 393 | while (enu.hasMoreElements()) { |
|---|
| 394 | newVector.addElement(enu.nextElement()); |
|---|
| 395 | } |
|---|
| 396 | return newVector.elements(); |
|---|
| 397 | } // listOptions |
|---|
| 398 | |
|---|
| 399 | /** |
|---|
| 400 | * Parses a given list of options. <p/> |
|---|
| 401 | * |
|---|
| 402 | <!-- options-start --> |
|---|
| 403 | * Valid options are: <p/> |
|---|
| 404 | * |
|---|
| 405 | * <pre> -P <nr of parents> |
|---|
| 406 | * Maximum number of parents</pre> |
|---|
| 407 | * |
|---|
| 408 | * <pre> -R |
|---|
| 409 | * Use arc reversal operation. |
|---|
| 410 | * (default false)</pre> |
|---|
| 411 | * |
|---|
| 412 | * <pre> -N |
|---|
| 413 | * Initial structure is empty (instead of Naive Bayes)</pre> |
|---|
| 414 | * |
|---|
| 415 | * <pre> -mbc |
|---|
| 416 | * Applies a Markov Blanket correction to the network structure, |
|---|
| 417 | * after a network structure is learned. This ensures that all |
|---|
| 418 | * nodes in the network are part of the Markov blanket of the |
|---|
| 419 | * classifier node.</pre> |
|---|
| 420 | * |
|---|
| 421 | * <pre> -S [LOO-CV|k-Fold-CV|Cumulative-CV] |
|---|
| 422 | * Score type (LOO-CV,k-Fold-CV,Cumulative-CV)</pre> |
|---|
| 423 | * |
|---|
| 424 | * <pre> -Q |
|---|
| 425 | * Use probabilistic or 0/1 scoring. |
|---|
| 426 | * (default probabilistic scoring)</pre> |
|---|
| 427 | * |
|---|
| 428 | <!-- options-end --> |
|---|
| 429 | * |
|---|
| 430 | * @param options the list of options as an array of strings |
|---|
| 431 | * @throws Exception if an option is not supported |
|---|
| 432 | */ |
|---|
| 433 | public void setOptions(String[] options) throws Exception { |
|---|
| 434 | setUseArcReversal(Utils.getFlag('R', options)); |
|---|
| 435 | |
|---|
| 436 | setInitAsNaiveBayes (!(Utils.getFlag('N', options))); |
|---|
| 437 | |
|---|
| 438 | String sMaxNrOfParents = Utils.getOption('P', options); |
|---|
| 439 | if (sMaxNrOfParents.length() != 0) { |
|---|
| 440 | setMaxNrOfParents(Integer.parseInt(sMaxNrOfParents)); |
|---|
| 441 | } else { |
|---|
| 442 | setMaxNrOfParents(100000); |
|---|
| 443 | } |
|---|
| 444 | |
|---|
| 445 | super.setOptions(options); |
|---|
| 446 | } // setOptions |
|---|
| 447 | |
|---|
| 448 | /** |
|---|
| 449 | * Gets the current settings of the search algorithm. |
|---|
| 450 | * |
|---|
| 451 | * @return an array of strings suitable for passing to setOptions |
|---|
| 452 | */ |
|---|
| 453 | public String[] getOptions() { |
|---|
| 454 | String[] superOptions = super.getOptions(); |
|---|
| 455 | String[] options = new String[7 + superOptions.length]; |
|---|
| 456 | int current = 0; |
|---|
| 457 | if (getUseArcReversal()) { |
|---|
| 458 | options[current++] = "-R"; |
|---|
| 459 | } |
|---|
| 460 | |
|---|
| 461 | if (!getInitAsNaiveBayes()) { |
|---|
| 462 | options[current++] = "-N"; |
|---|
| 463 | } |
|---|
| 464 | |
|---|
| 465 | options[current++] = "-P"; |
|---|
| 466 | options[current++] = "" + m_nMaxNrOfParents; |
|---|
| 467 | |
|---|
| 468 | // insert options from parent class |
|---|
| 469 | for (int iOption = 0; iOption < superOptions.length; iOption++) { |
|---|
| 470 | options[current++] = superOptions[iOption]; |
|---|
| 471 | } |
|---|
| 472 | |
|---|
| 473 | // Fill up rest with empty strings, not nulls! |
|---|
| 474 | while (current < options.length) { |
|---|
| 475 | options[current++] = ""; |
|---|
| 476 | } |
|---|
| 477 | return options; |
|---|
| 478 | } // getOptions |
|---|
| 479 | |
|---|
| 480 | /** |
|---|
| 481 | * Sets whether to init as naive bayes |
|---|
| 482 | * |
|---|
| 483 | * @param bInitAsNaiveBayes whether to init as naive bayes |
|---|
| 484 | */ |
|---|
| 485 | public void setInitAsNaiveBayes(boolean bInitAsNaiveBayes) { |
|---|
| 486 | m_bInitAsNaiveBayes = bInitAsNaiveBayes; |
|---|
| 487 | } |
|---|
| 488 | |
|---|
| 489 | /** |
|---|
| 490 | * Gets whether to init as naive bayes |
|---|
| 491 | * |
|---|
| 492 | * @return whether to init as naive bayes |
|---|
| 493 | */ |
|---|
| 494 | public boolean getInitAsNaiveBayes() { |
|---|
| 495 | return m_bInitAsNaiveBayes; |
|---|
| 496 | } |
|---|
| 497 | |
|---|
| 498 | /** get use the arc reversal operation |
|---|
| 499 | * @return whether the arc reversal operation should be used |
|---|
| 500 | */ |
|---|
| 501 | public boolean getUseArcReversal() { |
|---|
| 502 | return m_bUseArcReversal; |
|---|
| 503 | } // getUseArcReversal |
|---|
| 504 | |
|---|
| 505 | /** set use the arc reversal operation |
|---|
| 506 | * @param bUseArcReversal whether the arc reversal operation should be used |
|---|
| 507 | */ |
|---|
| 508 | public void setUseArcReversal(boolean bUseArcReversal) { |
|---|
| 509 | m_bUseArcReversal = bUseArcReversal; |
|---|
| 510 | } // setUseArcReversal |
|---|
| 511 | |
|---|
| 512 | /** |
|---|
| 513 | * This will return a string describing the search algorithm. |
|---|
| 514 | * @return The string. |
|---|
| 515 | */ |
|---|
| 516 | public String globalInfo() { |
|---|
| 517 | return "This Bayes Network learning algorithm uses a hill climbing algorithm " + |
|---|
| 518 | "adding, deleting and reversing arcs. The search is not restricted by an order " + |
|---|
| 519 | "on the variables (unlike K2). The difference with B and B2 is that this hill " + |
|---|
| 520 | "climber also considers arrows part of the naive Bayes structure for deletion."; |
|---|
| 521 | } // globalInfo |
|---|
| 522 | |
|---|
| 523 | /** |
|---|
| 524 | * @return a string to describe the Use Arc Reversal option. |
|---|
| 525 | */ |
|---|
| 526 | public String useArcReversalTipText() { |
|---|
| 527 | return "When set to true, the arc reversal operation is used in the search."; |
|---|
| 528 | } // useArcReversalTipText |
|---|
| 529 | |
|---|
| 530 | /** |
|---|
| 531 | * Returns the revision string. |
|---|
| 532 | * |
|---|
| 533 | * @return the revision |
|---|
| 534 | */ |
|---|
| 535 | public String getRevision() { |
|---|
| 536 | return RevisionUtils.extract("$Revision: 1.9 $"); |
|---|
| 537 | } |
|---|
| 538 | } // HillClimber |
|---|