[6152] | 1 | /* |
---|
| 2 | Copyright 2006 by Sean Luke |
---|
| 3 | Licensed under the Academic Free License version 3.0 |
---|
| 4 | See the file "LICENSE" for more information |
---|
| 5 | */ |
---|
| 6 | |
---|
| 7 | |
---|
| 8 | package ec.util; |
---|
| 9 | |
---|
| 10 | /* |
---|
| 11 | * RandomChoice.java |
---|
| 12 | * |
---|
| 13 | * Created: January 5, 2001 |
---|
| 14 | * By: Sean Luke |
---|
| 15 | * |
---|
| 16 | */ |
---|
| 17 | |
---|
| 18 | /** |
---|
| 19 | * RandomChoice organizes arrays of floats into distributions which can |
---|
| 20 | * be used to pick randomly from. You can provide three kinds of arrays: |
---|
| 21 | * |
---|
| 22 | <ul> |
---|
| 23 | <li> An array of floats |
---|
| 24 | <li> An array of doubles |
---|
| 25 | <li> An array of arbitrary objects, plus a RandomChoiceChooser which knows |
---|
| 26 | how to get and set the appropriate "float" value of objects in this array. |
---|
| 27 | </ul> |
---|
| 28 | * |
---|
| 29 | * <p>Before the RandomChoice can pick randomly from your array, it must first |
---|
| 30 | * organize it. It does this by doing the following. First, it normalizes the |
---|
| 31 | * values in the array. Then it modifies them to their sums. That is, each item i |
---|
| 32 | * in the array is set to the sum of the original values for items 0...i. If you |
---|
| 33 | * cannot allow your objects to be modified, then this is not the class for you. |
---|
| 34 | * |
---|
| 35 | * <p>An array is valid if (1) it has no negative values and (2) not all of its |
---|
| 36 | * values are zero. This RandomChoice code <i>should</i> (I hope) guarantee that |
---|
| 37 | * an element of zero probability is never returned. RandomChoice uses a binary |
---|
| 38 | * search to find your index, followed by linear probing (marching up or down |
---|
| 39 | * the list) to find the first non-zero probability item in the vacinity of that |
---|
| 40 | * index. As long as there are not a whole lot of zero-valued items in a row, |
---|
| 41 | * RandomChoice is efficient. |
---|
| 42 | * |
---|
| 43 | * You organize your array with organizeDistribution(). Then you can have the |
---|
| 44 | * RandomChoice pick random items from the array and return their indexes to you. |
---|
| 45 | * You do this by calling pickFromDistribution(), passing it a random floating |
---|
| 46 | * point value between 0.0 and 1.0. You call organizeDistribution() only once; |
---|
| 47 | * after which you may call pickFromDistribution() as many times as you like. |
---|
| 48 | * You should not modify the array thereafter. |
---|
| 49 | * |
---|
| 50 | * @author Sean Luke |
---|
| 51 | * @version 1.0 |
---|
| 52 | */ |
---|
| 53 | |
---|
| 54 | public class RandomChoice |
---|
| 55 | { |
---|
| 56 | |
---|
| 57 | /** Same as organizeDistribution(probabilities, <b>false</b>); */ |
---|
| 58 | public static void organizeDistribution(final float[] probabilities) |
---|
| 59 | { |
---|
| 60 | organizeDistribution(probabilities, false); |
---|
| 61 | } |
---|
| 62 | |
---|
| 63 | /** Normalizes probabilities, then converts them into continuing |
---|
| 64 | sums. This prepares them for being usable in pickFromDistribution. |
---|
| 65 | If the probabilities are all 0, then selection is uniform, unless allowAllZeros |
---|
| 66 | is false, in which case an ArithmeticException is thrown. If any of them are negative, |
---|
| 67 | or if the distribution is empty, then an ArithmeticException is thrown. |
---|
| 68 | For example, |
---|
| 69 | {0.6, 0.4, 0.2, 0.8} -> {0.3, 0.2, 0.1, 0.4} -> {0.3, 0.5, 0.6, 1.0} */ |
---|
| 70 | |
---|
| 71 | public static void organizeDistribution(final float[] probabilities, final boolean allowAllZeros) |
---|
| 72 | { |
---|
| 73 | // first normalize |
---|
| 74 | double sum=0.0; |
---|
| 75 | if (probabilities.length == 0) |
---|
| 76 | throw new ArithmeticException("Distribution has no elements"); |
---|
| 77 | |
---|
| 78 | for(int x=0;x<probabilities.length;x++) |
---|
| 79 | { |
---|
| 80 | if (probabilities[x]<0.0) |
---|
| 81 | throw new ArithmeticException("Distribution has negative probabilities"); |
---|
| 82 | sum += probabilities[x]; |
---|
| 83 | } |
---|
| 84 | |
---|
| 85 | if (sum==0.0) |
---|
| 86 | if (!allowAllZeros) |
---|
| 87 | throw new ArithmeticException("Distribution has all zero probabilities"); |
---|
| 88 | else |
---|
| 89 | { |
---|
| 90 | for(int x=0;x<probabilities.length;x++) |
---|
| 91 | probabilities[x] = 1.0f; |
---|
| 92 | sum = probabilities.length; |
---|
| 93 | } |
---|
| 94 | |
---|
| 95 | for(int x=0;x<probabilities.length;x++) |
---|
| 96 | probabilities[x] /= sum; |
---|
| 97 | |
---|
| 98 | // now sum |
---|
| 99 | sum=0.0; |
---|
| 100 | for(int x=0;x<probabilities.length;x++) |
---|
| 101 | { |
---|
| 102 | sum += probabilities[x]; |
---|
| 103 | probabilities[x] = (float)sum; |
---|
| 104 | } |
---|
| 105 | |
---|
| 106 | // now we need to work backwards setting 0 values |
---|
| 107 | int x; |
---|
| 108 | for(x=probabilities.length-1; x > 0; x--) |
---|
| 109 | if (probabilities[x]==probabilities[x-1]) // we're 0.0 |
---|
| 110 | probabilities[x] = 1.0f; |
---|
| 111 | else break; |
---|
| 112 | probabilities[x] = 1.0f; |
---|
| 113 | } |
---|
| 114 | |
---|
| 115 | /** Same as organizeDistribution(probabilities, <b>false</b>); */ |
---|
| 116 | public static void organizeDistribution(final double[] probabilities) |
---|
| 117 | { |
---|
| 118 | organizeDistribution(probabilities, false); |
---|
| 119 | } |
---|
| 120 | |
---|
| 121 | /** Normalizes probabilities, then converts them into continuing |
---|
| 122 | sums. This prepares them for being usable in pickFromDistribution. |
---|
| 123 | If the probabilities are all 0, then selection is uniform, unless allowAllZeros |
---|
| 124 | is false, in which case an ArithmeticException is thrown. If any of them are negative, |
---|
| 125 | or if the distribution is empty, then an ArithmeticException is thrown. |
---|
| 126 | For example, |
---|
| 127 | {0.6, 0.4, 0.2, 0.8} -> {0.3, 0.2, 0.1, 0.4} -> {0.3, 0.5, 0.6, 1.0} */ |
---|
| 128 | |
---|
| 129 | public static void organizeDistribution(final double[] probabilities, final boolean allowAllZeros) |
---|
| 130 | { |
---|
| 131 | // first normalize |
---|
| 132 | double sum=0.0; |
---|
| 133 | |
---|
| 134 | if (probabilities.length == 0) |
---|
| 135 | throw new ArithmeticException("Distribution has no elements"); |
---|
| 136 | |
---|
| 137 | for(int x=0;x<probabilities.length;x++) |
---|
| 138 | { |
---|
| 139 | if (probabilities[x]<0.0) |
---|
| 140 | throw new ArithmeticException("Distribution has negative probabilities"); |
---|
| 141 | sum += probabilities[x]; |
---|
| 142 | } |
---|
| 143 | |
---|
| 144 | if (sum==0.0) |
---|
| 145 | if (!allowAllZeros) |
---|
| 146 | throw new ArithmeticException("Distribution has all zero probabilities"); |
---|
| 147 | else |
---|
| 148 | { |
---|
| 149 | for(int x=0;x<probabilities.length;x++) |
---|
| 150 | probabilities[x] = 1.0; |
---|
| 151 | sum = probabilities.length; |
---|
| 152 | } |
---|
| 153 | |
---|
| 154 | for(int x=0;x<probabilities.length;x++) |
---|
| 155 | probabilities[x] /= sum; |
---|
| 156 | |
---|
| 157 | // now sum |
---|
| 158 | sum=0.0; |
---|
| 159 | for(int x=0;x<probabilities.length;x++) |
---|
| 160 | { |
---|
| 161 | sum += probabilities[x]; |
---|
| 162 | probabilities[x] = sum; |
---|
| 163 | } |
---|
| 164 | |
---|
| 165 | // now we need to work backwards setting 0 values |
---|
| 166 | int x; |
---|
| 167 | for(x=probabilities.length-1; x > 0; x--) |
---|
| 168 | if (probabilities[x]==probabilities[x-1]) // we're 0.0 |
---|
| 169 | probabilities[x] = 1.0; |
---|
| 170 | else break; |
---|
| 171 | probabilities[x] = 1.0; |
---|
| 172 | |
---|
| 173 | } |
---|
| 174 | |
---|
| 175 | /** Same as organizeDistribution(objs, chooser, <b>false</b>); */ |
---|
| 176 | public static void organizeDistribution(final Object[] objs, |
---|
| 177 | final RandomChoiceChooser chooser) |
---|
| 178 | { |
---|
| 179 | organizeDistribution(objs,chooser, false); |
---|
| 180 | } |
---|
| 181 | |
---|
| 182 | /** Normalizes the probabilities associated |
---|
| 183 | with an array of objects, then converts them into continuing |
---|
| 184 | sums. This prepares them for being usable in pickFromDistribution. |
---|
| 185 | If the probabilities are all 0, then selection is uniform, unless allowAllZeros |
---|
| 186 | is false, in which case an ArithmeticException is thrown. If any of them are negative, |
---|
| 187 | or if the distribution is empty, then an ArithmeticException is thrown. |
---|
| 188 | For example, |
---|
| 189 | {0.6, 0.4, 0.2, 0.8} -> {0.3, 0.2, 0.1, 0.4} -> {0.3, 0.5, 0.6, 1.0} |
---|
| 190 | The probabilities are retrieved and set using chooser.*/ |
---|
| 191 | |
---|
| 192 | public static void organizeDistribution(final Object[] objs, |
---|
| 193 | final RandomChoiceChooser chooser, final boolean allowAllZeros) |
---|
| 194 | { |
---|
| 195 | // first normalize |
---|
| 196 | double sum=0.0; |
---|
| 197 | |
---|
| 198 | if (objs.length == 0) |
---|
| 199 | throw new ArithmeticException("Distribution has no elements"); |
---|
| 200 | |
---|
| 201 | for(int x=0;x<objs.length;x++) |
---|
| 202 | { |
---|
| 203 | if (chooser.getProbability(objs[x])<0.0) |
---|
| 204 | throw new ArithmeticException("Distribution has negative probabilities"); |
---|
| 205 | sum += chooser.getProbability(objs[x]); |
---|
| 206 | } |
---|
| 207 | |
---|
| 208 | if (sum==0.0) |
---|
| 209 | if (!allowAllZeros) |
---|
| 210 | throw new ArithmeticException("Distribution has all zero probabilities"); |
---|
| 211 | else |
---|
| 212 | { |
---|
| 213 | for(int x=0;x<objs.length;x++) |
---|
| 214 | chooser.setProbability(objs[x], 1.0f); |
---|
| 215 | sum = objs.length; |
---|
| 216 | } |
---|
| 217 | |
---|
| 218 | for(int x=0;x<objs.length;x++) |
---|
| 219 | chooser.setProbability(objs[x], |
---|
| 220 | (float)(chooser.getProbability(objs[x]) / sum)); |
---|
| 221 | |
---|
| 222 | // now sum |
---|
| 223 | sum=0.0; |
---|
| 224 | for(int x=0;x<objs.length;x++) |
---|
| 225 | { |
---|
| 226 | sum += chooser.getProbability(objs[x]); |
---|
| 227 | chooser.setProbability(objs[x],(float)sum); |
---|
| 228 | } |
---|
| 229 | |
---|
| 230 | // now we need to work backwards setting 0 values |
---|
| 231 | int x; |
---|
| 232 | for(x=objs.length-1; x > 0; x--) |
---|
| 233 | if (chooser.getProbability(objs[x])== |
---|
| 234 | chooser.getProbability(objs[x-1])) // we're 0.0 |
---|
| 235 | chooser.setProbability(objs[x],1.0f); |
---|
| 236 | else break; |
---|
| 237 | chooser.setProbability(objs[x],1.0f); |
---|
| 238 | } |
---|
| 239 | |
---|
| 240 | /** Same as organizeDistribution(objs, chooser, <b>false</b>); */ |
---|
| 241 | public static void organizeDistribution(final Object[] objs, |
---|
| 242 | final RandomChoiceChooserD chooser) |
---|
| 243 | { |
---|
| 244 | organizeDistribution(objs,chooser, false); |
---|
| 245 | } |
---|
| 246 | |
---|
| 247 | /** Normalizes the probabilities associated |
---|
| 248 | with an array of objects, then converts them into continuing |
---|
| 249 | sums. This prepares them for being usable in pickFromDistribution. |
---|
| 250 | If the probabilities are all 0, then selection is uniform, unless allowAllZeros |
---|
| 251 | is false, in which case an ArithmeticException is thrown. If any of them are negative, |
---|
| 252 | or if the distribution is empty, then an ArithmeticException is thrown. |
---|
| 253 | For example, |
---|
| 254 | {0.6, 0.4, 0.2, 0.8} -> {0.3, 0.2, 0.1, 0.4} -> {0.3, 0.5, 0.6, 1.0} |
---|
| 255 | The probabilities are retrieved and set using chooser.*/ |
---|
| 256 | |
---|
| 257 | public static void organizeDistribution(final Object[] objs, |
---|
| 258 | final RandomChoiceChooserD chooser, final boolean allowAllZeros) |
---|
| 259 | { |
---|
| 260 | // first normalize |
---|
| 261 | double sum=0.0; |
---|
| 262 | |
---|
| 263 | if (objs.length == 0) |
---|
| 264 | throw new ArithmeticException("Distribution has no elements"); |
---|
| 265 | |
---|
| 266 | for(int x=0;x<objs.length;x++) |
---|
| 267 | { |
---|
| 268 | if (chooser.getProbability(objs[x])<0.0) |
---|
| 269 | throw new ArithmeticException("Distribution has negative probabilities"); |
---|
| 270 | sum += chooser.getProbability(objs[x]); |
---|
| 271 | } |
---|
| 272 | |
---|
| 273 | if (sum==0.0) |
---|
| 274 | if (!allowAllZeros) |
---|
| 275 | throw new ArithmeticException("Distribution has all zero probabilities"); |
---|
| 276 | else |
---|
| 277 | { |
---|
| 278 | for(int x=0;x<objs.length;x++) |
---|
| 279 | chooser.setProbability(objs[x], 1.0); |
---|
| 280 | sum = objs.length; |
---|
| 281 | } |
---|
| 282 | |
---|
| 283 | for(int x=0;x<objs.length;x++) |
---|
| 284 | chooser.setProbability(objs[x], |
---|
| 285 | (double)(chooser.getProbability(objs[x]) / sum)); |
---|
| 286 | |
---|
| 287 | // now sum |
---|
| 288 | sum=0.0; |
---|
| 289 | for(int x=0;x<objs.length;x++) |
---|
| 290 | { |
---|
| 291 | sum += chooser.getProbability(objs[x]); |
---|
| 292 | chooser.setProbability(objs[x],(double)sum); |
---|
| 293 | } |
---|
| 294 | |
---|
| 295 | // now we need to work backwards setting 0 values |
---|
| 296 | int x; |
---|
| 297 | for(x=objs.length-1; x > 0; x--) |
---|
| 298 | if (chooser.getProbability(objs[x])== |
---|
| 299 | chooser.getProbability(objs[x-1])) // we're 0.0 |
---|
| 300 | chooser.setProbability(objs[x],1.0); |
---|
| 301 | else break; |
---|
| 302 | chooser.setProbability(objs[x],1.0); |
---|
| 303 | } |
---|
| 304 | |
---|
| 305 | |
---|
| 306 | // allows us to have zero-probability values |
---|
| 307 | private static final int exemptZeroes(final float[] probabilities, int index) |
---|
| 308 | { |
---|
| 309 | //System.out.println(index); |
---|
| 310 | if (probabilities[index]==0.0f) // I need to scan forward because I'm in a left-trail |
---|
| 311 | // scan forward |
---|
| 312 | { while(index < probabilities.length-1 && probabilities[index]==0.0f) index++; } |
---|
| 313 | else |
---|
| 314 | // scan backwards |
---|
| 315 | { while(index > 0 && probabilities[index]==probabilities[index-1]) index--; } |
---|
| 316 | return index; |
---|
| 317 | } |
---|
| 318 | |
---|
| 319 | // allows us to have zero-probability values |
---|
| 320 | private static final int exemptZeroes(final double[] probabilities, int index) |
---|
| 321 | { |
---|
| 322 | //System.out.println(index); |
---|
| 323 | if (probabilities[index]==0.0) // I need to scan forward because I'm in a left-trail |
---|
| 324 | // scan forward |
---|
| 325 | { while(index < probabilities.length-1 && probabilities[index]==0.0) index++; } |
---|
| 326 | else |
---|
| 327 | // scan backwards |
---|
| 328 | { while(index > 0 && probabilities[index]==probabilities[index-1]) index--; } |
---|
| 329 | return index; |
---|
| 330 | } |
---|
| 331 | |
---|
| 332 | |
---|
| 333 | // allows us to have zero-probability values |
---|
| 334 | private static final int exemptZeroes(final Object[] objs, |
---|
| 335 | final RandomChoiceChooser chooser, int index) |
---|
| 336 | { |
---|
| 337 | //System.out.println(index); |
---|
| 338 | if (chooser.getProbability(objs[index])==0.0f) // I need to scan forward because I'm in a left-trail |
---|
| 339 | // scan forward |
---|
| 340 | { while(index < objs.length-1 && chooser.getProbability(objs[index])==0.0f) index++; } |
---|
| 341 | else |
---|
| 342 | // scan backwards |
---|
| 343 | { while(index > 0 && chooser.getProbability(objs[index])== |
---|
| 344 | chooser.getProbability(objs[index-1])) index--; } |
---|
| 345 | return index; |
---|
| 346 | } |
---|
| 347 | |
---|
| 348 | |
---|
| 349 | // allows us to have zero-probability values |
---|
| 350 | private static final int exemptZeroes(final Object[] objs, |
---|
| 351 | final RandomChoiceChooserD chooser, int index) |
---|
| 352 | { |
---|
| 353 | //System.out.println(index); |
---|
| 354 | if (chooser.getProbability(objs[index])==0.0) // I need to scan forward because I'm in a left-trail |
---|
| 355 | // scan forward |
---|
| 356 | { while(index < objs.length-1 && chooser.getProbability(objs[index])==0.0) index++; } |
---|
| 357 | else |
---|
| 358 | // scan backwards |
---|
| 359 | { while(index > 0 && chooser.getProbability(objs[index])== |
---|
| 360 | chooser.getProbability(objs[index-1])) index--; } |
---|
| 361 | return index; |
---|
| 362 | } |
---|
| 363 | |
---|
| 364 | |
---|
| 365 | public static final int CHECKBOUNDARY = 8; |
---|
| 366 | |
---|
| 367 | /** Picks a random item from an array of probabilities, |
---|
| 368 | normalized and summed as follows: For example, |
---|
| 369 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 370 | they should get normalized and summed by the outside owners |
---|
| 371 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < CHECKBOUNDARY, |
---|
| 372 | then a linear search is used, else a binary search is used. */ |
---|
| 373 | |
---|
| 374 | public static int pickFromDistribution(final float[] probabilities, |
---|
| 375 | final float prob) |
---|
| 376 | { |
---|
| 377 | return pickFromDistribution(probabilities, prob, CHECKBOUNDARY); |
---|
| 378 | } |
---|
| 379 | |
---|
| 380 | |
---|
| 381 | /** Picks a random item from an array of probabilities, |
---|
| 382 | normalized and summed as follows: For example, |
---|
| 383 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 384 | they should get normalized and summed by the outside owners |
---|
| 385 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < checkboundary, |
---|
| 386 | then a linear search is used, else a binary search is used. |
---|
| 387 | @deprecated |
---|
| 388 | */ |
---|
| 389 | public static int pickFromDistribution(final float[] probabilities, |
---|
| 390 | final float prob, final int checkboundary) |
---|
| 391 | { |
---|
| 392 | if (prob<0.0f || prob>1.0f) |
---|
| 393 | throw new ArithmeticException("Invalid probability for pickFromDistribution (must be 0.0<=x<=1.0)"); |
---|
| 394 | else if (probabilities.length==1) // quick |
---|
| 395 | return 0; |
---|
| 396 | else if (probabilities.length<checkboundary) |
---|
| 397 | { |
---|
| 398 | // simple linear scan |
---|
| 399 | for(int x=0;x<probabilities.length-1;x++) |
---|
| 400 | if (probabilities[x]>prob) |
---|
| 401 | return exemptZeroes(probabilities,x); |
---|
| 402 | return exemptZeroes(probabilities,probabilities.length-1); |
---|
| 403 | } |
---|
| 404 | else |
---|
| 405 | { |
---|
| 406 | // binary search |
---|
| 407 | int top = probabilities.length-1; |
---|
| 408 | int bottom = 0; |
---|
| 409 | int cur; |
---|
| 410 | |
---|
| 411 | while(top!=bottom) |
---|
| 412 | { |
---|
| 413 | cur = (top + bottom) / 2; // integer division |
---|
| 414 | |
---|
| 415 | if (probabilities[cur] > prob) |
---|
| 416 | if (cur==0 || probabilities[cur-1] <= prob) |
---|
| 417 | return exemptZeroes(probabilities,cur); |
---|
| 418 | else // step down |
---|
| 419 | top = cur; |
---|
| 420 | else if (cur==probabilities.length-1) // oops |
---|
| 421 | return exemptZeroes(probabilities,cur); |
---|
| 422 | else if (bottom==cur) // step up |
---|
| 423 | bottom++; // (8 + 9)/2 = 8 |
---|
| 424 | else |
---|
| 425 | bottom = cur; // (8 + 10) / 2 = 9 |
---|
| 426 | } |
---|
| 427 | return exemptZeroes(probabilities,bottom); // oops |
---|
| 428 | } |
---|
| 429 | } |
---|
| 430 | |
---|
| 431 | |
---|
| 432 | |
---|
| 433 | /** Picks a random item from an array of probabilities, |
---|
| 434 | normalized and summed as follows: For example, |
---|
| 435 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 436 | they should get normalized and summed by the outside owners |
---|
| 437 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < CHECKBOUNDARY, |
---|
| 438 | then a linear search is used, else a binary search is used. */ |
---|
| 439 | |
---|
| 440 | public static int pickFromDistribution(final double[] probabilities, |
---|
| 441 | final double prob) |
---|
| 442 | { |
---|
| 443 | return pickFromDistribution(probabilities, prob, CHECKBOUNDARY); |
---|
| 444 | } |
---|
| 445 | |
---|
| 446 | |
---|
| 447 | /** Picks a random item from an array of probabilities, |
---|
| 448 | normalized and summed as follows: For example, |
---|
| 449 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 450 | they should get normalized and summed by the outside owners |
---|
| 451 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < checkboundary, |
---|
| 452 | then a linear search is used, else a binary search is used. |
---|
| 453 | @deprecated |
---|
| 454 | */ |
---|
| 455 | |
---|
| 456 | public static int pickFromDistribution(final double[] probabilities, |
---|
| 457 | final double prob, final int checkboundary) |
---|
| 458 | { |
---|
| 459 | if (prob<0.0 || prob>1.0) |
---|
| 460 | throw new ArithmeticException("Invalid probability for pickFromDistribution (must be 0.0<=x<=1.0)"); |
---|
| 461 | if (probabilities.length==1) // quick |
---|
| 462 | return 0; |
---|
| 463 | else if (probabilities.length<checkboundary) |
---|
| 464 | { |
---|
| 465 | // simple linear scan |
---|
| 466 | for(int x=0;x<probabilities.length-1;x++) |
---|
| 467 | if (probabilities[x]>prob) |
---|
| 468 | return exemptZeroes(probabilities,x); |
---|
| 469 | return exemptZeroes(probabilities,probabilities.length-1); |
---|
| 470 | } |
---|
| 471 | else |
---|
| 472 | { |
---|
| 473 | // binary search |
---|
| 474 | int top = probabilities.length-1; |
---|
| 475 | int bottom = 0; |
---|
| 476 | int cur; |
---|
| 477 | |
---|
| 478 | while(top!=bottom) |
---|
| 479 | { |
---|
| 480 | cur = (top + bottom) / 2; // integer division |
---|
| 481 | |
---|
| 482 | if (probabilities[cur] > prob) |
---|
| 483 | if (cur==0 || probabilities[cur-1] <= prob) |
---|
| 484 | return exemptZeroes(probabilities,cur); |
---|
| 485 | else // step down |
---|
| 486 | top = cur; |
---|
| 487 | else if (cur==probabilities.length-1) // oops |
---|
| 488 | return exemptZeroes(probabilities,cur); |
---|
| 489 | else if (bottom==cur) // step up |
---|
| 490 | bottom++; // (8 + 9)/2 = 8 |
---|
| 491 | else |
---|
| 492 | bottom = cur; // (8 + 10) / 2 = 9 |
---|
| 493 | } |
---|
| 494 | return exemptZeroes(probabilities,bottom); // oops |
---|
| 495 | } |
---|
| 496 | } |
---|
| 497 | |
---|
| 498 | |
---|
| 499 | /** Picks a random item from an array of objects, each with an |
---|
| 500 | associated probability that is accessed by taking an object |
---|
| 501 | and passing it to chooser.getProbability(obj). The objects' |
---|
| 502 | probabilities are |
---|
| 503 | normalized and summed as follows: For example, |
---|
| 504 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 505 | they should get normalized and summed by the outside owners |
---|
| 506 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < CHECKBOUNDARY, |
---|
| 507 | then a linear search is used, else a binary search is used. */ |
---|
| 508 | |
---|
| 509 | public static int pickFromDistribution(final Object[] objs, |
---|
| 510 | final RandomChoiceChooser chooser, |
---|
| 511 | final float prob) |
---|
| 512 | { |
---|
| 513 | return pickFromDistribution(objs, chooser, prob, CHECKBOUNDARY); |
---|
| 514 | } |
---|
| 515 | |
---|
| 516 | /** Picks a random item from an array of objects, each with an |
---|
| 517 | associated probability that is accessed by taking an object |
---|
| 518 | and passing it to chooser.getProbability(obj). The objects' |
---|
| 519 | probabilities are |
---|
| 520 | normalized and summed as follows: For example, |
---|
| 521 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 522 | they should get normalized and summed by the outside owners |
---|
| 523 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < checkboundary, |
---|
| 524 | then a linear search is used, else a binary search is used. |
---|
| 525 | @deprecated |
---|
| 526 | */ |
---|
| 527 | |
---|
| 528 | public static int pickFromDistribution(final Object[] objs, |
---|
| 529 | final RandomChoiceChooser chooser, |
---|
| 530 | final float prob, final int checkboundary) |
---|
| 531 | { |
---|
| 532 | if (prob<0.0f || prob>1.0f) |
---|
| 533 | throw new ArithmeticException("Invalid probability for pickFromDistribution (must be 0.0<=x<=1.0)"); |
---|
| 534 | if (objs.length==1) // quick |
---|
| 535 | return 0; |
---|
| 536 | else if (objs.length<checkboundary) |
---|
| 537 | { |
---|
| 538 | // simple linear scan |
---|
| 539 | for(int x=0;x<objs.length-1;x++) |
---|
| 540 | if (chooser.getProbability(objs[x]) >prob) |
---|
| 541 | return exemptZeroes(objs,chooser,x); |
---|
| 542 | return exemptZeroes(objs,chooser,objs.length-1); |
---|
| 543 | } |
---|
| 544 | else |
---|
| 545 | { |
---|
| 546 | // binary search |
---|
| 547 | int top = objs.length-1; |
---|
| 548 | int bottom = 0; |
---|
| 549 | int cur; |
---|
| 550 | |
---|
| 551 | while(top!=bottom) |
---|
| 552 | { |
---|
| 553 | cur = (top + bottom) / 2; // integer division |
---|
| 554 | |
---|
| 555 | if (chooser.getProbability(objs[cur]) > prob) |
---|
| 556 | if (cur==0 || chooser.getProbability(objs[cur-1]) <= prob) |
---|
| 557 | return exemptZeroes(objs,chooser,cur); |
---|
| 558 | else // step down |
---|
| 559 | top = cur; |
---|
| 560 | else if (cur==objs.length-1) // oops |
---|
| 561 | return exemptZeroes(objs,chooser,cur); |
---|
| 562 | else if (bottom==cur) // step up |
---|
| 563 | bottom++; // (8 + 9)/2 = 8 |
---|
| 564 | else |
---|
| 565 | bottom = cur; // (8 + 10) / 2 = 9 |
---|
| 566 | } |
---|
| 567 | return exemptZeroes(objs,chooser,bottom); // oops |
---|
| 568 | } |
---|
| 569 | } |
---|
| 570 | |
---|
| 571 | |
---|
| 572 | /** Picks a random item from an array of objects, each with an |
---|
| 573 | associated probability that is accessed by taking an object |
---|
| 574 | and passing it to chooser.getProbability(obj). The objects' |
---|
| 575 | probabilities are |
---|
| 576 | normalized and summed as follows: For example, |
---|
| 577 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 578 | they should get normalized and summed by the outside owners |
---|
| 579 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < CHECKBOUNDARY, |
---|
| 580 | then a linear search is used, else a binary search is used. */ |
---|
| 581 | |
---|
| 582 | public static int pickFromDistribution(final Object[] objs, |
---|
| 583 | final RandomChoiceChooserD chooser, |
---|
| 584 | final double prob) |
---|
| 585 | { |
---|
| 586 | return pickFromDistribution(objs, chooser, prob, CHECKBOUNDARY); |
---|
| 587 | } |
---|
| 588 | |
---|
| 589 | /** Picks a random item from an array of objects, each with an |
---|
| 590 | associated probability that is accessed by taking an object |
---|
| 591 | and passing it to chooser.getProbability(obj). The objects' |
---|
| 592 | probabilities are |
---|
| 593 | normalized and summed as follows: For example, |
---|
| 594 | if four probabilities are {0.3, 0.2, 0.1, 0.4}, then |
---|
| 595 | they should get normalized and summed by the outside owners |
---|
| 596 | as: {0.3, 0.5, 0.6, 1.0}. If probabilities.length < checkboundary, |
---|
| 597 | then a linear search is used, else a binary search is used. |
---|
| 598 | @deprecated |
---|
| 599 | */ |
---|
| 600 | |
---|
| 601 | public static int pickFromDistribution(final Object[] objs, |
---|
| 602 | final RandomChoiceChooserD chooser, |
---|
| 603 | final double prob, final int checkboundary) |
---|
| 604 | { |
---|
| 605 | if (prob<0.0 || prob>1.0) |
---|
| 606 | throw new ArithmeticException("Invalid probability for pickFromDistribution (must be 0.0<=x<=1.0)"); |
---|
| 607 | if (objs.length==1) // quick |
---|
| 608 | return 0; |
---|
| 609 | else if (objs.length<checkboundary) |
---|
| 610 | { |
---|
| 611 | // simple linear scan |
---|
| 612 | for(int x=0;x<objs.length-1;x++) |
---|
| 613 | if (chooser.getProbability(objs[x]) >prob) |
---|
| 614 | return exemptZeroes(objs,chooser,x); |
---|
| 615 | return exemptZeroes(objs,chooser,objs.length-1); |
---|
| 616 | } |
---|
| 617 | else |
---|
| 618 | { |
---|
| 619 | // binary search |
---|
| 620 | int top = objs.length-1; |
---|
| 621 | int bottom = 0; |
---|
| 622 | int cur; |
---|
| 623 | |
---|
| 624 | while(top!=bottom) |
---|
| 625 | { |
---|
| 626 | cur = (top + bottom) / 2; // integer division |
---|
| 627 | |
---|
| 628 | if (chooser.getProbability(objs[cur]) > prob) |
---|
| 629 | if (cur==0 || chooser.getProbability(objs[cur-1]) <= prob) |
---|
| 630 | return exemptZeroes(objs,chooser,cur); |
---|
| 631 | else // step down |
---|
| 632 | top = cur; |
---|
| 633 | else if (cur==objs.length-1) // oops |
---|
| 634 | return exemptZeroes(objs,chooser,cur); |
---|
| 635 | else if (bottom==cur) // step up |
---|
| 636 | bottom++; // (8 + 9)/2 = 8 |
---|
| 637 | else |
---|
| 638 | bottom = cur; // (8 + 10) / 2 = 9 |
---|
| 639 | } |
---|
| 640 | return exemptZeroes(objs,chooser,bottom); // oops |
---|
| 641 | } |
---|
| 642 | } |
---|
| 643 | |
---|
| 644 | } |
---|
| 645 | |
---|
| 646 | |
---|
| 647 | |
---|
| 648 | |
---|