Scippy

SCIP

Solving Constraint Integer Programs

branch_relpscost.c
Go to the documentation of this file.
1 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
2 /* */
3 /* This file is part of the program and library */
4 /* SCIP --- Solving Constraint Integer Programs */
5 /* */
6 /* Copyright (c) 2002-2024 Zuse Institute Berlin (ZIB) */
7 /* */
8 /* Licensed under the Apache License, Version 2.0 (the "License"); */
9 /* you may not use this file except in compliance with the License. */
10 /* You may obtain a copy of the License at */
11 /* */
12 /* http://www.apache.org/licenses/LICENSE-2.0 */
13 /* */
14 /* Unless required by applicable law or agreed to in writing, software */
15 /* distributed under the License is distributed on an "AS IS" BASIS, */
16 /* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. */
17 /* See the License for the specific language governing permissions and */
18 /* limitations under the License. */
19 /* */
20 /* You should have received a copy of the Apache-2.0 license */
21 /* along with SCIP; see the file LICENSE. If not visit scipopt.org. */
22 /* */
23 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
24 
25 /**@file branch_relpscost.c
26  * @ingroup DEFPLUGINS_BRANCH
27  * @brief reliable pseudo costs branching rule
28  * @author Tobias Achterberg
29  * @author Timo Berthold
30  * @author Gerald Gamrath
31  * @author Marc Pfetsch
32  */
33 
34 /*---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8----+----9----+----0----+----1----+----2*/
35 
36 #include "blockmemshell/memory.h"
37 #include "scip/branch_relpscost.h"
38 #include "scip/treemodel.h"
39 #include "scip/cons_and.h"
40 #include "scip/pub_branch.h"
41 #include "scip/pub_cons.h"
42 #include "scip/pub_message.h"
43 #include "scip/pub_misc.h"
44 #include "scip/pub_sol.h"
45 #include "scip/pub_tree.h"
46 #include "scip/pub_var.h"
47 #include "scip/scip_branch.h"
48 #include "scip/scip_cons.h"
49 #include "scip/scip_general.h"
50 #include "scip/scip_lp.h"
51 #include "scip/scip_mem.h"
52 #include "scip/scip_message.h"
53 #include "scip/scip_nlp.h"
54 #include "scip/scip_numerics.h"
55 #include "scip/scip_param.h"
56 #include "scip/scip_prob.h"
57 #include "scip/scip_randnumgen.h"
58 #include "scip/scip_sol.h"
59 #include "scip/scip_solvingstats.h"
60 #include "scip/scip_tree.h"
61 #include "scip/scip_var.h"
62 #include "scip/prop_symmetry.h"
63 #include "scip/symmetry.h"
64 #include <string.h>
65 
66 #define BRANCHRULE_NAME "relpscost"
67 #define BRANCHRULE_DESC "reliability branching on pseudo cost values"
68 #define BRANCHRULE_PRIORITY 10000
69 #define BRANCHRULE_MAXDEPTH -1
70 #define BRANCHRULE_MAXBOUNDDIST 1.0
71 
72 #define DEFAULT_CONFLICTWEIGHT 0.01 /**< weight in score calculations for conflict score */
73 #define DEFAULT_CONFLENGTHWEIGHT 0.0 /**< weight in score calculations for conflict length score*/
74 #define DEFAULT_INFERENCEWEIGHT 0.0001 /**< weight in score calculations for inference score */
75 #define DEFAULT_CUTOFFWEIGHT 0.0001 /**< weight in score calculations for cutoff score */
76 #define DEFAULT_GMIAVGEFFWEIGHT 0.0 /**< weight in score calculations of average GMI cut normed efficacies */
77 #define DEFAULT_GMILASTEFFWEIGHT 0.00001 /**< weight in score calculations of last GMI cut normed efficacy */
78 #define DEFAULT_PSCOSTWEIGHT 1.0 /**< weight in score calculations for pseudo cost score */
79 #define DEFAULT_NLSCOREWEIGHT 0.1 /**< weight in score calculations for nlcount score */
80 #define DEFAULT_MINRELIABLE 1.0 /**< minimal value for minimum pseudo cost size to regard pseudo cost value as reliable */
81 #define DEFAULT_MAXRELIABLE 5.0 /**< maximal value for minimum pseudo cost size to regard pseudo cost value as reliable */
82 #define DEFAULT_SBITERQUOT 0.5 /**< maximal fraction of strong branching LP iterations compared to normal iterations */
83 #define DEFAULT_SBITEROFS 100000 /**< additional number of allowed strong branching LP iterations */
84 #define DEFAULT_MAXLOOKAHEAD 9 /**< maximal number of further variables evaluated without better score */
85 #define DEFAULT_INITCAND 100 /**< maximal number of candidates initialized with strong branching per node */
86 #define DEFAULT_INITITER 0 /**< iteration limit for strong branching initialization of pseudo cost entries (0: auto) */
87 #define DEFAULT_MAXBDCHGS 5 /**< maximal number of bound tightenings before the node is reevaluated (-1: unlimited) */
88 #define DEFAULT_MAXPROPROUNDS -2 /**< maximum number of propagation rounds to be performed during strong branching
89  * before solving the LP (-1: no limit, -2: parameter settings) */
90 #define DEFAULT_PROBINGBOUNDS TRUE /**< should valid bounds be identified in a probing-like fashion during strong
91  * branching (only with propagation)? */
92 #define DEFAULT_USERELERRORFORRELIABILITY FALSE /**< should reliability be based on relative errors? */
93 #define DEFAULT_LOWERRORTOL 0.05 /**< lowest tolerance beneath which relative errors are reliable */
94 #define DEFAULT_HIGHERRORTOL 1.0 /**< highest tolerance beneath which relative errors are reliable */
95 #define DEFAULT_USEHYPTESTFORRELIABILITY FALSE /**< should the strong branching decision be based on a hypothesis test? */
96 #define DEFAULT_USEDYNAMICCONFIDENCE FALSE /**< should the confidence level be adjusted dynamically? */
97 #define DEFAULT_STORESEMIINITCOSTS FALSE /**< should strong branching result be considered for pseudo costs if the other direction was infeasible? */
98 #define DEFAULT_USESBLOCALINFO FALSE /**< should the scoring function use only local cutoff and inference information obtained for strong branching candidates? */
99 #define DEFAULT_CONFIDENCELEVEL 2 /**< The confidence level for statistical methods, between 0 (Min) and 4 (Max). */
100 #define DEFAULT_SKIPBADINITCANDS TRUE /**< should branching rule skip candidates that have a low probability to be
101  * better than the best strong-branching or pseudo-candidate? */
102 #define DEFAULT_STARTRANDSEED 5 /**< start random seed for random number generation */
103 #define DEFAULT_RANDINITORDER FALSE /**< should slight perturbation of scores be used to break ties in the prior scores? */
104 #define DEFAULT_USESMALLWEIGHTSITLIM FALSE /**< should smaller weights be used for pseudo cost updates after hitting the LP iteration limit? */
105 #define DEFAULT_DYNAMICWEIGHTS TRUE /**< should the weights of the branching rule be adjusted dynamically during solving based
106  * infeasible and objective leaf counters? */
107 #define DEFAULT_DEGENERACYAWARE 1 /**< should degeneracy be taken into account to update weights and skip strong branching? (0: off, 1: after root, 2: always)*/
109 /* symmetry handling */
110 #define DEFAULT_FILTERCANDSSYM FALSE /**< Use symmetry to filter branching candidates? */
111 #define DEFAULT_TRANSSYMPSCOST FALSE /**< Transfer pscost information to symmetric variables if filtering is performed? */
112 
113 /** branching rule data */
114 struct SCIP_BranchruleData
115 {
116  SCIP_Real conflictweight; /**< weight in score calculations for conflict score */
117  SCIP_Real conflengthweight; /**< weight in score calculations for conflict length score */
118  SCIP_Real inferenceweight; /**< weight in score calculations for inference score */
119  SCIP_Real cutoffweight; /**< weight in score calculations for cutoff score */
120  SCIP_Real gmiavgeffweight; /**< weight in score calculations of average GMI normed cut efficacies */
121  SCIP_Real gmilasteffweight; /**< weight in score calculations of last GMI cut normalized efficacy */
122  SCIP_Real pscostweight; /**< weight in score calculations for pseudo cost score */
123  SCIP_Real nlscoreweight; /**< weight in score calculations for nlcount score */
124  SCIP_Real minreliable; /**< minimal value for minimum pseudo cost size to regard pseudo cost value as reliable */
125  SCIP_Real maxreliable; /**< maximal value for minimum pseudo cost size to regard pseudo cost value as reliable */
126  SCIP_Real sbiterquot; /**< maximal fraction of strong branching LP iterations compared to normal iterations */
127  int sbiterofs; /**< additional number of allowed strong branching LP iterations */
128  int maxlookahead; /**< maximal number of further variables evaluated without better score */
129  int initcand; /**< maximal number of candidates initialized with strong branching per node */
130  int inititer; /**< iteration limit for strong branching initialization of pseudo cost entries (0: auto) */
131  int maxbdchgs; /**< maximal number of bound tightenings before the node is reevaluated (-1: unlimited) */
132  int maxproprounds; /**< maximum number of propagation rounds to be performed during strong branching
133  * before solving the LP (-1: no limit, -2: parameter settings) */
134  SCIP_Bool probingbounds; /**< should valid bounds be identified in a probing-like fashion during strong
135  * branching (only with propagation)? */
136  SCIP_Bool userelerrorforreliability; /**< should reliability be based on relative errors? */
137  SCIP_Real lowerrortol; /**< lowest tolerance beneath which relative errors are reliable */
138  SCIP_Real higherrortol; /**< highest tolerance beneath which relative errors are reliable */
139  SCIP_Bool usehyptestforreliability; /**< should the strong branching decision be based on a hypothesis test? */
140  SCIP_Bool usedynamicconfidence; /**< should the confidence level be adjusted dynamically? */
141  SCIP_Bool storesemiinitcosts; /**< should strong branching result be considered for pseudo costs if the
142  * other direction was infeasible? */
143  SCIP_Bool usesblocalinfo; /**< should the scoring function disregard cutoffs for variable if sb-lookahead was feasible ? */
144  SCIP_Bool skipbadinitcands; /**< should branching rule skip candidates that have a low probability to be
145  * better than the best strong-branching or pseudo-candidate? */
146  SCIP_Bool dynamicweights; /**< should the weights of the branching rule be adjusted dynamically during
147  * solving based on objective and infeasible leaf counters? */
148  int degeneracyaware; /**< should degeneracy be taken into account to update weights and skip strong branching? (0: off, 1: after root, 2: always) */
149  int confidencelevel; /**< The confidence level for statistical methods, between 0 (Min) and 4 (Max). */
150  int* nlcount; /**< array to store nonlinear count values */
151  int nlcountsize; /**< length of nlcount array */
152  int nlcountmax; /**< maximum entry in nlcount array or 1 if NULL */
153  SCIP_Bool randinitorder; /**< should slight perturbation of scores be used to break ties in the prior scores? */
154  SCIP_RANDNUMGEN* randnumgen; /**< random number generator */
155  int startrandseed; /**< start random seed for random number generation */
156  SCIP_Bool usesmallweightsitlim; /**< should smaller weights be used for pseudo cost updates after hitting the LP iteration limit? */
157  SCIP_TREEMODEL* treemodel; /**< Parameters for the Treemodel branching rules */
158 
159  /* for symmetry */
160  SCIP_Bool filtercandssym; /**< Use symmetry to filter branching candidates? */
161  SCIP_Bool transsympscost; /**< Transfer pscost information to symmetric variables? */
162 
163  SCIP_Bool nosymmetry; /**< No symmetry present? */
164  int* orbits; /**< array of non-trivial orbits */
165  int* orbitbegins; /**< array containing begin positions of new orbits in orbits array */
166  int norbits; /**< pointer to number of orbits currently stored in orbits */
167  int* varorbitmap; /**< array for storing indices of the containing orbit for each variable */
168  int* orbitrep; /**< representative variable of each orbit */
169  SCIP_VAR** permvars; /**< variables on which permutations act */
170  int npermvars; /**< number of variables for permutations */
171  SCIP_HASHMAP* permvarmap; /**< map of variables to indices in permvars array */
172 };
173 
174 /*
175  * local methods
176  */
177 
178 /** initialize orbits */
179 static
181  SCIP* scip, /**< SCIP data structure */
182  SCIP_BRANCHRULEDATA* branchruledata /**< branching rule data */
183  )
184 {
185  int** permstrans = NULL;
186  int* components = NULL;
187  int* componentbegins = NULL;
188  int* vartocomponent = NULL;
189  int ncomponents = 0;
190  int nperms = -1;
191 
192  assert( scip != NULL );
193  assert( branchruledata != NULL );
194  assert( branchruledata->filtercandssym );
195 
196  /* exit if no symmetry or orbits already available */
197  if( branchruledata->nosymmetry || branchruledata->orbits != NULL )
198  return SCIP_OKAY;
199 
200  assert( branchruledata->orbitbegins == NULL );
201  assert( branchruledata->varorbitmap == NULL );
202  assert( branchruledata->orbitrep == NULL );
203 
204  /* obtain symmetry including permutations */
205  SCIP_CALL( SCIPgetSymmetry(scip, &branchruledata->npermvars, &branchruledata->permvars, &branchruledata->permvarmap,
206  &nperms, NULL, &permstrans, NULL, NULL, &components, &componentbegins, &vartocomponent, &ncomponents) );
207 
208  /* turn off symmetry handling if there is no symmetry or the number of variables is not equal */
209  if( nperms <= 0 || branchruledata->npermvars != SCIPgetNVars(scip) )
210  {
211  branchruledata->nosymmetry = TRUE;
212  return SCIP_OKAY;
213  }
214  assert( branchruledata->permvars != NULL );
215  assert( branchruledata->permvarmap != NULL );
216  assert( branchruledata->npermvars > 0 );
217  assert( permstrans != NULL );
218  assert( components != NULL );
219  assert( componentbegins != NULL );
220  assert( vartocomponent != NULL );
221  assert( ncomponents > 0 );
222 
223  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &branchruledata->orbits, branchruledata->npermvars) );
224  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &branchruledata->orbitbegins, branchruledata->npermvars) );
225  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &branchruledata->varorbitmap, branchruledata->npermvars) );
226  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &branchruledata->orbitrep, branchruledata->npermvars) );
227 
228  /* Compute orbits on all variables, since this might help for branching and this computation is only done once. */
229  SCIP_CALL( SCIPcomputeOrbitsComponentsSym(scip, branchruledata->npermvars, permstrans, nperms,
230  components, componentbegins, vartocomponent, ncomponents,
231  branchruledata->orbits, branchruledata->orbitbegins, &branchruledata->norbits, branchruledata->varorbitmap) );
232  assert( branchruledata->norbits < branchruledata->npermvars );
233 
234  return SCIP_OKAY;
235 }
236 
237 /** filter out symmetric variables from branching variables */
238 static
240  SCIP* scip, /**< SCIP data structure */
241  SCIP_BRANCHRULEDATA* branchruledata, /**< branching rule data */
242  SCIP_VAR** origbranchcands, /**< original branching candidates */
243  SCIP_Real* origbranchcandssol, /**< original solution value for the branching candidates */
244  SCIP_Real* origbranchcandsfrac,/**< original fractional part of the branching candidates */
245  int norigbranchcands, /**< original number of branching candidates */
246  SCIP_VAR** branchcands, /**< branching candidates */
247  SCIP_Real* branchcandssol, /**< solution value for the branching candidates */
248  SCIP_Real* branchcandsfrac, /**< fractional part of the branching candidates */
249  int* branchorbitidx, /**< array of indices of orbit of branching candidates */
250  int* nbranchcands /**< pointer to store number of branching candidates */
251  )
252 {
253  int i;
254 
255  assert( scip != NULL );
256  assert( branchruledata != NULL );
257  assert( origbranchcands != NULL );
258  assert( origbranchcandssol != NULL );
259  assert( origbranchcandsfrac != NULL );
260  assert( branchcands != NULL );
261  assert( branchcandssol != NULL );
262  assert( branchcandsfrac != NULL );
263  assert( nbranchcands != NULL );
264 
265  assert( ! branchruledata->nosymmetry );
266  assert( branchruledata->orbitbegins != NULL );
267  assert( branchruledata->orbits != NULL );
268  assert( branchruledata->permvarmap != NULL );
269  assert( branchruledata->varorbitmap != NULL );
270  assert( branchruledata->orbitrep != NULL );
271  assert( branchruledata->norbits < branchruledata->npermvars );
272 
273  /* init representatives (used to see whether variable is the first in an orbit) */
274  for( i = 0; i < branchruledata->norbits; ++i )
275  branchruledata->orbitrep[i] = -1;
276 
277  /* loop through branching variables, determine orbit and whether they are the first ones */
278  *nbranchcands = 0;
279  for( i = 0; i < norigbranchcands; ++i )
280  {
281  int orbitidx = -1;
282  int varidx;
283 
284  varidx = SCIPhashmapGetImageInt(branchruledata->permvarmap, (void*) origbranchcands[i]);
285  if( varidx != INT_MAX )
286  {
287  assert( 0 <= varidx && varidx < branchruledata->npermvars );
288  orbitidx = branchruledata->varorbitmap[varidx];
289  }
290  assert( -1 <= orbitidx && orbitidx < branchruledata->norbits );
291 
292  /* Check whether the variable is not present (can happen if variable was added after computing symmetries or is in
293  * a singleton orbit). */
294  if( orbitidx == -1 )
295  {
296  branchcands[*nbranchcands] = origbranchcands[i];
297  branchcandssol[*nbranchcands] = origbranchcandssol[i];
298  branchcandsfrac[*nbranchcands] = origbranchcandsfrac[i];
299  branchorbitidx[*nbranchcands] = -1;
300  ++(*nbranchcands);
301  }
302  else if( branchruledata->orbitrep[orbitidx] == -1 )
303  {
304  /* if variable is the first in a nontrivial orbit */
305  assert( 0 <= varidx && varidx < branchruledata->npermvars );
306  branchruledata->orbitrep[orbitidx] = varidx;
307  branchcands[*nbranchcands] = origbranchcands[i];
308  branchcandssol[*nbranchcands] = origbranchcandssol[i];
309  branchcandsfrac[*nbranchcands] = origbranchcandsfrac[i];
310  branchorbitidx[*nbranchcands] = orbitidx;
311  ++(*nbranchcands);
312  }
313  }
314 
315  SCIPdebugMsg(scip, "Filtered out %d variables by symmetry.\n", norigbranchcands - *nbranchcands);
316 
317  return SCIP_OKAY;
318 }
319 
320 /** updates the pseudo costs of the given variable and all its symmetric variables */
321 static
323  SCIP* scip, /**< SCIP data structure */
324  SCIP_BRANCHRULEDATA* branchruledata, /**< branching rule data */
325  SCIP_VAR* branchvar, /**< branching variable candidate */
326  int* branchorbitidx, /**< array of orbit indices */
327  int branchvaridx, /**< index of variable in branchorbitidx */
328  SCIP_Real solvaldelta, /**< difference of variable's new LP value - old LP value */
329  SCIP_Real objdelta, /**< difference of new LP's objective value - old LP's objective value */
330  SCIP_Real weight /**< weight in (0,1] of this update in pseudo cost sum */
331  )
332 {
333  int orbitidx;
334  int j;
335 
336  assert( scip != NULL );
337  assert( branchruledata != NULL );
338 
339  if( branchruledata->nosymmetry || ! branchruledata->transsympscost || branchorbitidx == NULL )
340  {
341  /* use original update function */
342  SCIP_CALL( SCIPupdateVarPseudocost(scip, branchvar, solvaldelta, objdelta, weight) );
343  return SCIP_OKAY;
344  }
345 
346  assert( branchruledata->orbitbegins != NULL );
347  assert( branchruledata->orbits != NULL );
348  assert( 0 <= branchvaridx && branchvaridx < branchruledata->npermvars );
349 
350  orbitidx = branchorbitidx[branchvaridx];
351  if( orbitidx < 0 )
352  {
353  /* only update given variable */
354  SCIP_CALL( SCIPupdateVarPseudocost(scip, branchvar, solvaldelta, objdelta, weight) );
355  return SCIP_OKAY;
356  }
357  assert( 0 <= orbitidx && orbitidx < branchruledata->norbits );
358 
359  /* loop through orbit containing variable and update pseudo costs for all variables */
360  for( j = branchruledata->orbitbegins[orbitidx]; j < branchruledata->orbitbegins[orbitidx+1]; ++j )
361  {
362  SCIP_VAR* var;
363  int idx;
364 
365  idx = branchruledata->orbits[j];
366  assert( 0 <= idx && idx < branchruledata->npermvars );
367 
368  var = branchruledata->permvars[idx];
369  assert( var != NULL );
370 
371  if( SCIPvarIsActive(var) )
372  {
373  SCIP_CALL( SCIPupdateVarPseudocost(scip, var, solvaldelta, objdelta, weight) );
374  }
375  }
376 
377  return SCIP_OKAY;
378 }
379 
380 /**! [SnippetCodeStyleStaticAsserts] */
381 
382 /** return probindex of variable or corresponding active variable (if negated or aggregated) or -1 (if multiaggregated) */
383 static
385  SCIP* scip, /**< SCIP data structure */
386  SCIP_VAR* var, /**< binary variable */
387  int* probindex /**< buffer to store probindex */
388  )
389 {
390  assert(scip != NULL);
391  assert(var != NULL);
392  assert(SCIPvarIsBinary(var));
393  assert(probindex != NULL);
394 
395 /**! [SnippetCodeStyleStaticAsserts] */
396 
397  *probindex = SCIPvarGetProbindex(var);
398 
399  /* if variable is not active, try to find active representative */
400  if( *probindex == -1 )
401  {
402  SCIP_VAR* repvar;
403  SCIP_Bool negated;
404 
405  SCIP_CALL( SCIPgetBinvarRepresentative(scip, var, &repvar, &negated) );
406  assert(repvar != NULL);
407  assert(SCIPvarGetStatus(repvar) != SCIP_VARSTATUS_FIXED);
408 
409  if( SCIPvarIsActive(repvar) )
410  *probindex = SCIPvarGetProbindex(repvar);
411  else if( SCIPvarIsNegated(repvar) )
412  *probindex = SCIPvarGetProbindex(SCIPvarGetNegationVar(repvar));
413  }
414 
415  return SCIP_OKAY;
416 }
417 
418 /**! [SnippetCodeStyleDeclaration] */
419 
420 /** counts number of nonlinear constraints in which each variable appears */
421 static
423  SCIP* scip, /**< SCIP data structure */
424  int* nlcount, /**< pointer to array for storing count values */
425  int nlcountsize, /**< buffer for storing length of nlcount array */
426  int* nlcountmax /**< buffer for storing maximum value in nlcount array */
427  )
428 {
429  SCIP_CONSHDLR* andconshdlr;
430  SCIP_VAR** vars;
431  int nvars;
432  int i;
433 
434 /**! [SnippetCodeStyleDeclaration] */
435 
436  assert(scip != NULL);
437  assert(nlcount != NULL);
438  assert(nlcountmax != NULL);
439 
440  SCIP_CALL( SCIPgetVarsData(scip, &vars, &nvars, NULL, NULL, NULL, NULL) );
441  assert(nlcountsize >= nvars);
442 
443  /* get nonlinearity for constraints in NLP */
444  if( SCIPisNLPConstructed(scip) )
445  {
446  assert(SCIPgetNNLPVars(scip) == nvars);
447  SCIP_CALL( SCIPgetNLPVarsNonlinearity(scip, nlcount) );
448  }
449  else
450  {
451  BMSclearMemoryArray(nlcount, nvars);
452  }
453 
454  /* increase counters for and constraints */
455  andconshdlr = SCIPfindConshdlr(scip, "and");
456  if( andconshdlr != NULL )
457  {
458  int c;
459 
460  for( c = 0; c < SCIPconshdlrGetNActiveConss(andconshdlr); c++ )
461  {
462  SCIP_CONS* andcons;
463  SCIP_VAR** andvars;
464  SCIP_VAR* andres;
465  int probindex;
466  int nandvars;
467  int v;
468 
469  /* get constraint and variables */
470  andcons = SCIPconshdlrGetConss(andconshdlr)[c];
471  nandvars = SCIPgetNVarsAnd(scip, andcons);
472  andvars = SCIPgetVarsAnd(scip, andcons);
473  andres = SCIPgetResultantAnd(scip, andcons);
474 
475  probindex = -1;
476 
477  /**! [SnippetCodeStyleIfFor] */
478 
479  for( v = 0; v < nandvars; v++ )
480  {
481  /* don't rely on the and conshdlr removing fixed variables
482  * @todo fix the and conshdlr in that respect
483  */
484  if( SCIPvarGetStatus(andvars[v]) != SCIP_VARSTATUS_FIXED )
485  {
486  SCIP_CALL( binvarGetActiveProbindex(scip, andvars[v], &probindex) );
487  if( probindex >= 0 )
488  nlcount[probindex]++;
489  }
490  }
491 
492  /**! [SnippetCodeStyleIfFor] */
493 
494  SCIP_CALL( binvarGetActiveProbindex(scip, andres, &probindex) );
495  if( probindex >= 0 )
496  nlcount[probindex]++;
497  }
498  }
499 
500  /* compute maximum count value */
501  *nlcountmax = 1;
502  for( i = 0; i < nvars; i++ )
503  {
504  if( *nlcountmax < nlcount[i] )
505  *nlcountmax = nlcount[i];
506  }
507 
508  return SCIP_OKAY;
509 }
510 
511 static
513  SCIP* scip, /**< SCIP data structure */
514  SCIP_BRANCHRULEDATA* branchruledata /**< branching rule data */
515  )
516 {
517  int nvars;
518 
519  assert(scip != NULL);
520  assert(branchruledata != NULL);
521 
522  nvars = SCIPgetNVars(scip);
523 
524  /**@todo test whether we want to apply this as if problem has only and constraints */
525  /**@todo update changes in and constraints */
526  if( branchruledata->nlscoreweight > 0.0 ) /* && SCIPisNLPConstructed(scip) */
527  {
528  if( branchruledata->nlcount == NULL )
529  {
530  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &branchruledata->nlcount, nvars) );
531  branchruledata->nlcountsize = nvars;
532 
533  SCIP_CALL( countNonlinearities(scip, branchruledata->nlcount, branchruledata->nlcountsize, &branchruledata->nlcountmax) );
534  }
535  else if( branchruledata->nlcountsize < nvars )
536  {
537  SCIP_CALL( SCIPreallocBlockMemoryArray(scip, &branchruledata->nlcount, branchruledata->nlcountsize, nvars) );
538  /**@todo should we update nlcounts for new variables? */
539  BMSclearMemoryArray(&(branchruledata->nlcount[branchruledata->nlcountsize]), nvars - branchruledata->nlcountsize); /*lint !e866*/
540  branchruledata->nlcountsize = nvars;
541  }
542  assert(branchruledata->nlcount != NULL);
543  assert(branchruledata->nlcountsize == nvars);
544  assert(branchruledata->nlcountmax >= 1);
545  }
546  else
547  {
548  SCIPfreeBlockMemoryArrayNull(scip, &branchruledata->nlcount, branchruledata->nlcountsize);
549  branchruledata->nlcountsize = 0;
550  branchruledata->nlcountmax = 1;
551  }
552 
553  return SCIP_OKAY;
554 }
555 
556 
557 /** calculates nlscore value between 0 and 1 */
558 static
560  SCIP* scip, /**< SCIP data structure */
561  int* nlcount, /**< array to store count values */
562  int nlcountmax, /**< maximum value in nlcount array */
563  int probindex /**< index of branching candidate */
564  )
565 {
566  if( nlcountmax >= 1 && nlcount != NULL )
567  {
568  SCIP_Real nlscore;
569 
570  assert(scip != NULL);
571  assert(probindex >= 0);
572  assert(probindex < SCIPgetNVars(scip));
573 
574  nlscore = nlcount[probindex] / (SCIP_Real)nlcountmax;
575 
576  assert(nlscore >= 0.0);
577  assert(nlscore <= 1.0);
578  return nlscore;
579  }
580  else
581  return 0.0;
582 }
583 
584 /** calculates an overall score value for the given individual score values */
585 static
587  SCIP* scip, /**< SCIP data structure */
588  SCIP_BRANCHRULEDATA* branchruledata, /**< branching rule data */
589  SCIP_Real conflictscore, /**< conflict score of current variable */
590  SCIP_Real avgconflictscore, /**< average conflict score */
591  SCIP_Real conflengthscore, /**< conflict length score of current variable */
592  SCIP_Real avgconflengthscore, /**< average conflict length score */
593  SCIP_Real inferencescore, /**< inference score of current variable */
594  SCIP_Real avginferencescore, /**< average inference score */
595  SCIP_Real cutoffscore, /**< cutoff score of current variable */
596  SCIP_Real avgcutoffscore, /**< average cutoff score */
597  SCIP_Real gmieffscore, /**< normalized-eff of avg GMI cuts from row when var was frac and basic */
598  SCIP_Real lastgmieffscore, /**< last normalized gmieffscore when var was frac and basic */
599  SCIP_Real pscostscore, /**< pscost score of current variable */
600  SCIP_Real avgpscostscore, /**< average pscost score */
601  SCIP_Real nlscore, /**< nonlinear score of current variable between 0 and 1 */
602  SCIP_Real frac, /**< fractional value of variable in current solution */
603  SCIP_Real degeneracyfactor /**< factor to apply because of degeneracy */
604  )
605 {
606  SCIP_Real score;
607  SCIP_Real dynamicfactor;
608 
609  assert(branchruledata != NULL);
610  assert(0.0 < frac && frac < 1.0);
611 
612  if( branchruledata->dynamicweights )
613  {
614  dynamicfactor = (SCIPgetNInfeasibleLeaves(scip) + 1.0) / (SCIPgetNObjlimLeaves(scip) + 1.0);
615  }
616  else
617  dynamicfactor = 1.0;
618 
619  dynamicfactor *= degeneracyfactor;
620 
621  score = dynamicfactor * (branchruledata->conflictweight * (1.0 - 1.0/(1.0+conflictscore/avgconflictscore))
622  + branchruledata->conflengthweight * (1.0 - 1.0/(1.0+conflengthscore/avgconflengthscore))
623  + branchruledata->inferenceweight * (1.0 - 1.0/(1.0+inferencescore/avginferencescore))
624  + branchruledata->cutoffweight * (1.0 - 1.0/(1.0+cutoffscore/avgcutoffscore))
625  + branchruledata->gmiavgeffweight * gmieffscore + branchruledata->gmilasteffweight * lastgmieffscore)
626  + branchruledata->pscostweight / dynamicfactor * (1.0 - 1.0/(1.0+pscostscore/avgpscostscore))
627  + branchruledata->nlscoreweight * nlscore;
628 
629  /* avoid close to integral variables */
630  if( MIN(frac, 1.0 - frac) < 10.0 * SCIPfeastol(scip) )
631  score *= 1e-6;
632 
633  return score;
634 }
635 
636 /** adds given index and direction to bound change arrays */
637 static
639  SCIP* scip, /**< SCIP data structure */
640  int** bdchginds, /**< pointer to bound change index array */
641  SCIP_BOUNDTYPE** bdchgtypes, /**< pointer to bound change types array */
642  SCIP_Real** bdchgbounds, /**< pointer to bound change new bounds array */
643  int* nbdchgs, /**< pointer to number of bound changes */
644  int ind, /**< index to store in bound change index array */
645  SCIP_BOUNDTYPE type, /**< type of the bound change to store in bound change type array */
646  SCIP_Real bound /**< new bound to store in bound change new bounds array */
647  )
648 {
649  assert(bdchginds != NULL);
650  assert(bdchgtypes != NULL);
651  assert(bdchgbounds != NULL);
652  assert(nbdchgs != NULL);
653 
654  SCIP_CALL( SCIPreallocBufferArray(scip, bdchginds, (*nbdchgs) + 1) );
655  SCIP_CALL( SCIPreallocBufferArray(scip, bdchgtypes, (*nbdchgs) + 1) );
656  SCIP_CALL( SCIPreallocBufferArray(scip, bdchgbounds, (*nbdchgs) + 1) );
657  assert(*bdchginds != NULL);
658  assert(*bdchgtypes != NULL);
659  assert(*bdchgbounds != NULL);
660 
661  (*bdchginds)[*nbdchgs] = ind;
662  (*bdchgtypes)[*nbdchgs] = type;
663  (*bdchgbounds)[*nbdchgs] = bound;
664  (*nbdchgs)++;
665 
666  return SCIP_OKAY;
667 }
668 
669 /** frees bound change arrays */
670 static
671 void freeBdchgs(
672  SCIP* scip, /**< SCIP data structure */
673  int** bdchginds, /**< pointer to bound change index array */
674  SCIP_BOUNDTYPE** bdchgtypes, /**< pointer to bound change types array */
675  SCIP_Real** bdchgbounds, /**< pointer to bound change new bounds array */
676  int* nbdchgs /**< pointer to number of bound changes */
677  )
678 {
679  assert(bdchginds != NULL);
680  assert(bdchgtypes != NULL);
681  assert(bdchgbounds != NULL);
682  assert(nbdchgs != NULL);
683 
684  SCIPfreeBufferArrayNull(scip, bdchgbounds);
685  SCIPfreeBufferArrayNull(scip, bdchgtypes);
686  SCIPfreeBufferArrayNull(scip, bdchginds);
687  *nbdchgs = 0;
688 }
689 
690 /** applies bound changes stored in bound change arrays */
691 static
693  SCIP* scip, /**< SCIP data structure */
694  SCIP_VAR** vars, /**< problem variables */
695  int* bdchginds, /**< bound change index array */
696  SCIP_BOUNDTYPE* bdchgtypes, /**< bound change types array */
697  SCIP_Real* bdchgbounds, /**< bound change new bound array */
698  int nbdchgs, /**< number of bound changes */
699  SCIP_RESULT* result /**< result pointer */
700  )
701 {
702 #ifndef NDEBUG
703  SCIP_BRANCHRULE* branchrule;
704  SCIP_BRANCHRULEDATA* branchruledata;
705 #endif
706  SCIP_Bool infeasible;
707  SCIP_Bool tightened;
708  int i;
709 
710  assert(vars != NULL);
711 
712 #ifndef NDEBUG
713  /* find branching rule */
714  branchrule = SCIPfindBranchrule(scip, BRANCHRULE_NAME);
715  assert(branchrule != NULL);
716 
717  /* get branching rule data */
718  branchruledata = SCIPbranchruleGetData(branchrule);
719  assert(branchruledata != NULL);
720 #endif
721 
722  SCIPdebugMsg(scip, "applying %d bound changes\n", nbdchgs);
723 
724  for( i = 0; i < nbdchgs; ++i )
725  {
726  int v;
727 
728  v = bdchginds[i];
729 
730  SCIPdebugMsg(scip, " -> <%s> [%g,%g]\n",
731  SCIPvarGetName(vars[v]), SCIPvarGetLbLocal(vars[v]), SCIPvarGetUbLocal(vars[v]));
732 
733  if( bdchgtypes[i] == SCIP_BOUNDTYPE_LOWER )
734  {
735  /* change lower bound of variable to given bound */
736  SCIP_CALL( SCIPtightenVarLb(scip, vars[v], bdchgbounds[i], TRUE, &infeasible, &tightened) );
737  if( infeasible )
738  {
739  *result = SCIP_CUTOFF;
740  return SCIP_OKAY;
741  }
742 
743  /* if we did propagation, the bound change might already have been added */
744  assert(tightened || (branchruledata->maxproprounds != 0));
745  }
746  else
747  {
748  assert(bdchgtypes[i] == SCIP_BOUNDTYPE_UPPER);
749 
750  /* change upper bound of variable to given bound */
751  SCIP_CALL( SCIPtightenVarUb(scip, vars[v], bdchgbounds[i], TRUE, &infeasible, &tightened) );
752  if( infeasible )
753  {
754  *result = SCIP_CUTOFF;
755  return SCIP_OKAY;
756  }
757 
758  /* if we did propagation, the bound change might already have been added */
759  assert(tightened || (branchruledata->maxproprounds != 0));
760  }
761  SCIPdebugMsg(scip, " -> [%g,%g]\n", SCIPvarGetLbLocal(vars[v]), SCIPvarGetUbLocal(vars[v]));
762  }
763 
764  return SCIP_OKAY;
765 }
766 
767 /** execute reliability pseudo cost branching */
768 static
770  SCIP* scip, /**< SCIP data structure */
771  SCIP_BRANCHRULE* branchrule, /**< branching rule */
772  SCIP_VAR** branchcands, /**< branching candidates */
773  SCIP_Real* branchcandssol, /**< solution value for the branching candidates */
774  SCIP_Real* branchcandsfrac, /**< fractional part of the branching candidates */
775  int* branchorbitidx, /**< indices of orbit (or NULL) */
776  int nbranchcands, /**< number of branching candidates */
777  SCIP_Bool executebranch, /**< execute a branching step or run probing only */
778  SCIP_RESULT* result /**< pointer to the result of the execution */
779  )
780 { /*lint --e{715}*/
781  SCIP_BRANCHRULEDATA* branchruledata;
782  SCIP_Real lpobjval;
783  SCIP_Real bestsbdown;
784  SCIP_Real bestsbup;
785  SCIP_Real provedbound;
786  SCIP_Bool bestsbdownvalid;
787  SCIP_Bool bestsbupvalid;
788  SCIP_Bool bestsbdowncutoff;
789  SCIP_Bool bestsbupcutoff;
790  SCIP_Bool bestisstrongbranch;
791  SCIP_Bool allcolsinlp;
792  SCIP_Bool exactsolve;
793  int ninitcands;
794  int bestcand;
795 
796  /* remember which variables strong branching is performed on, and the
797  * recorded lp bound changes that are observed */
798  SCIP_Bool* sbvars = NULL;
799  SCIP_Real* sbdown = NULL;
800  SCIP_Real* sbup = NULL;
801  SCIP_Bool* sbdownvalid = NULL;
802  SCIP_Bool* sbupvalid = NULL;
803 
804  *result = SCIP_DIDNOTRUN;
805 
806  assert(SCIPgetLPSolstat(scip) == SCIP_LPSOLSTAT_OPTIMAL);
807 
808  /* get branching rule data */
809  branchruledata = SCIPbranchruleGetData(branchrule);
810  assert(branchruledata != NULL);
811 
812  /* get current LP objective bound of the local sub problem and global cutoff bound */
813  lpobjval = SCIPgetLPObjval(scip);
814 
815  /* check, if we want to solve the problem exactly, meaning that strong branching information is not useful
816  * for cutting off sub problems and improving lower bounds of children
817  */
818  exactsolve = SCIPisExactSolve(scip);
819 
820  /* check, if all existing columns are in LP, and thus the strong branching results give lower bounds */
821  allcolsinlp = SCIPallColsInLP(scip);
822 
823  bestcand = -1;
824  bestisstrongbranch = FALSE;
825  bestsbdown = SCIP_INVALID;
826  bestsbup = SCIP_INVALID;
827  bestsbdownvalid = FALSE;
828  bestsbupvalid = FALSE;
829  bestsbdowncutoff = FALSE;
830  bestsbupcutoff = FALSE;
831  provedbound = lpobjval;
832 
833  /* Allocate memory to store the lp bounds of the up and down children
834  * for those of the variables that we performed sb on
835  */
836  SCIP_CALL( SCIPallocBufferArray(scip, &sbdown, nbranchcands) );
837  SCIP_CALL( SCIPallocBufferArray(scip, &sbup, nbranchcands) );
838  SCIP_CALL( SCIPallocBufferArray(scip, &sbdownvalid, nbranchcands) );
839  SCIP_CALL( SCIPallocBufferArray(scip, &sbupvalid, nbranchcands) );
840  SCIP_CALL( SCIPallocBufferArray(scip, &sbvars, nbranchcands) );
841 
842  if( nbranchcands == 1 )
843  {
844  /* only one candidate: nothing has to be done */
845  bestcand = 0;
846  SCIPdebug(ninitcands = 0);
847  sbvars[0] = FALSE;
848  }
849  else
850  {
851  SCIP_VAR** vars;
852  int* initcands;
853  SCIP_Real* initcandscores;
854  SCIP_Real* newlbs = NULL;
855  SCIP_Real* newubs = NULL;
856  SCIP_Real* mingains = NULL;
857  SCIP_Real* maxgains = NULL;
858  /* scores computed from pseudocost branching */
859  SCIP_Real* scores = NULL;
860  SCIP_Real* scoresfrompc = NULL;
861  SCIP_Real* scoresfromothers = NULL;
862  int* bdchginds;
863  SCIP_BOUNDTYPE* bdchgtypes;
864  SCIP_Real* bdchgbounds;
865  int maxninitcands;
866  int nuninitcands;
867  int nbdchgs;
868  int nbdconflicts;
869  SCIP_Real avgconflictscore;
870  SCIP_Real avgconflengthscore;
871  SCIP_Real avginferencescore;
872  SCIP_Real avgcutoffscore;
873  SCIP_Real avgpscostscore;
874  SCIP_Real bestpsscore;
875  SCIP_Real bestpsfracscore;
876  SCIP_Real bestpsdomainscore;
877  SCIP_Real bestsbscore;
878  SCIP_Real bestuninitsbscore;
879  SCIP_Real bestsbfracscore;
880  SCIP_Real bestsbdomainscore;
881  SCIP_Real prio;
882  SCIP_Real reliable;
883  SCIP_Real maxlookahead;
884  SCIP_Real lookahead;
885  SCIP_Real relerrorthreshold;
886  SCIP_Bool initstrongbranching;
887  SCIP_Bool propagate;
888  SCIP_Bool probingbounds;
889  SCIP_Longint nodenum;
890  SCIP_Longint nlpiterationsquot;
891  SCIP_Longint nsblpiterations;
892  SCIP_Longint maxnsblpiterations;
893  int bestsolidx;
894  int maxbdchgs;
895  int bestpscand;
896  int bestsbcand;
897  int bestuninitsbcand;
898  int inititer;
899  int nvars;
900  int i;
901  int c;
902  SCIP_CONFIDENCELEVEL clevel;
903  SCIP_Real degeneracyfactor = 1.0;
904 
905  /* get LP degeneracy information and compute a factor to change weighting of pseudo cost score vs. other scores */
906  if( branchruledata->degeneracyaware > 0 && (SCIPgetDepth(scip) > 0 || branchruledata->degeneracyaware > 1) )
907  {
908  SCIP_Real degeneracy;
909  SCIP_Real varconsratio;
910 
911  /* get LP degeneracy information */
912  SCIP_CALL( SCIPgetLPDualDegeneracy(scip, &degeneracy, &varconsratio) );
913 
914  assert(degeneracy >= 0.0);
915  assert(degeneracy <= 1.0);
916  assert(varconsratio >= 1.0);
917 
918  /* increase factor for a degeneracy >= 80% */
919  if( degeneracy >= 0.8 )
920  {
921  degeneracy = 10.0 * (degeneracy - 0.7);
922  degeneracyfactor = degeneracyfactor * pow(10.0,degeneracy);
923  }
924  /* increase factor for a variable-constraint ratio >= 2.0 */
925  if( varconsratio >= 2.0 )
926  {
927  degeneracyfactor *= 10.0 * varconsratio;
928  }
929  }
930 
931  vars = SCIPgetVars(scip);
932  nvars = SCIPgetNVars(scip);
933 
934  bestsolidx = SCIPgetBestSol(scip) == NULL ? -1 : SCIPsolGetIndex(SCIPgetBestSol(scip));
935 
936  /* get average conflict, inference, and pseudocost scores */
937  avgconflictscore = SCIPgetAvgConflictScore(scip);
938  avgconflictscore = MAX(avgconflictscore, 0.1);
939  avgconflengthscore = SCIPgetAvgConflictlengthScore(scip);
940  avgconflengthscore = MAX(avgconflengthscore, 0.1);
941  avginferencescore = SCIPgetAvgInferenceScore(scip);
942  avginferencescore = MAX(avginferencescore, 0.1);
943  avgcutoffscore = SCIPgetAvgCutoffScore(scip);
944  avgcutoffscore = MAX(avgcutoffscore, 0.1);
945  avgpscostscore = SCIPgetAvgPseudocostScore(scip);
946  avgpscostscore = MAX(avgpscostscore, 0.1);
947 
948  /* get nonlinear counts according to parameters */
949  SCIP_CALL( branchruledataEnsureNlcount(scip, branchruledata) );
950 
951  initstrongbranching = FALSE;
952 
953  /* check whether propagation should be performed */
954  propagate = (branchruledata->maxproprounds != 0);
955 
956  /* check whether valid bounds should be identified in probing-like fashion */
957  probingbounds = propagate && branchruledata->probingbounds;
958 
959  /* get maximal number of candidates to initialize with strong branching; if the current solutions is not basic,
960  * we cannot warmstart the simplex algorithm and therefore don't initialize any candidates
961  */
962  maxninitcands = MIN(nbranchcands, branchruledata->initcand);
963  if( !SCIPisLPSolBasic(scip) )
964  maxninitcands = 0;
965 
966  /* calculate maximal number of strong branching LP iterations; if we used too many, don't apply strong branching
967  * any more; also, if degeneracy is too high, don't run strong branching at this node
968  */
969  nlpiterationsquot = (SCIP_Longint)(branchruledata->sbiterquot * SCIPgetNNodeLPIterations(scip));
970  maxnsblpiterations = nlpiterationsquot + branchruledata->sbiterofs + SCIPgetNRootStrongbranchLPIterations(scip);
971  nsblpiterations = SCIPgetNStrongbranchLPIterations(scip);
972  if( nsblpiterations > maxnsblpiterations || degeneracyfactor >= 10.0 )
973  maxninitcands = 0;
974 
975  /* get buffer for storing the unreliable candidates */
976  SCIP_CALL( SCIPallocBufferArray(scip, &initcands, maxninitcands+1) ); /* allocate one additional slot for convenience */
977  SCIP_CALL( SCIPallocBufferArray(scip, &initcandscores, maxninitcands+1) );
978  ninitcands = 0;
979 
980  /* Allocate memory for the down and up gains, and the computed pseudocost scores */
981  SCIP_CALL( SCIPallocBufferArray(scip, &mingains, nbranchcands) );
982  SCIP_CALL( SCIPallocBufferArray(scip, &maxgains, nbranchcands) );
983  SCIP_CALL( SCIPallocBufferArray(scip, &scores, nbranchcands) );
984  SCIP_CALL( SCIPallocBufferArray(scip, &scoresfrompc, nbranchcands) );
985  SCIP_CALL( SCIPallocBufferArray(scip, &scoresfromothers, nbranchcands) );
986 
987  /* get current node number */
988  nodenum = SCIPgetNNodes(scip);
989 
990  /* initialize bound change arrays */
991  bdchginds = NULL;
992  bdchgtypes = NULL;
993  bdchgbounds = NULL;
994  nbdchgs = 0;
995  nbdconflicts = 0;
996  maxbdchgs = branchruledata->maxbdchgs;
997  if( maxbdchgs == -1 )
998  maxbdchgs = INT_MAX;
999 
1000  /* calculate value used as reliability */
1001  prio = (maxnsblpiterations - nsblpiterations)/(nsblpiterations + 1.0);
1002  prio = MIN(prio, 1.0);
1003  prio = MAX(prio, (nlpiterationsquot - nsblpiterations)/(nsblpiterations + 1.0));
1004  reliable = (1.0-prio) * branchruledata->minreliable + prio * branchruledata->maxreliable;
1005 
1006  /* calculate the threshold for the relative error in the same way; low tolerance is more strict than higher tolerance */
1007  relerrorthreshold = (1.0 - prio) * branchruledata->higherrortol + prio * branchruledata->lowerrortol;
1008 
1009  clevel = (SCIP_CONFIDENCELEVEL)branchruledata->confidencelevel;
1010  /* determine the confidence level for hypothesis testing based on value of prio */
1011  if( branchruledata->usedynamicconfidence )
1012  {
1013  /* with decreasing priority, use a less strict confidence level */
1014  if( prio >= 0.9 )
1015  clevel = SCIP_CONFIDENCELEVEL_MAX;
1016  else if( prio >= 0.7 )
1017  clevel = SCIP_CONFIDENCELEVEL_HIGH;
1018  else if( prio >= 0.5 )
1019  clevel = SCIP_CONFIDENCELEVEL_MEDIUM;
1020  else if( prio >= 0.3 )
1021  clevel = SCIP_CONFIDENCELEVEL_LOW;
1022  else
1023  clevel = SCIP_CONFIDENCELEVEL_MIN;
1024  }
1025 
1026  /* search for the best pseudo cost candidate, while remembering unreliable candidates in a sorted buffer */
1027  nuninitcands = 0;
1028  bestpscand = -1;
1029  bestpsscore = -SCIPinfinity(scip);
1030  bestpsfracscore = -SCIPinfinity(scip);
1031  bestpsdomainscore = -SCIPinfinity(scip);
1032 
1033  /* search for the best candidate first */
1034  if( branchruledata->usehyptestforreliability )
1035  {
1036  for( c = 0; c < nbranchcands; ++c )
1037  {
1038  SCIP_Real conflictscore;
1039  SCIP_Real conflengthscore;
1040  SCIP_Real inferencescore;
1041  SCIP_Real cutoffscore;
1042  SCIP_Real gmieffscore;
1043  SCIP_Real lastgmieffscore;
1044  SCIP_Real pscostscore;
1045  SCIP_Real nlscore;
1046  SCIP_Real score;
1047 
1048  conflictscore = SCIPgetVarConflictScore(scip, branchcands[c]);
1049  conflengthscore = SCIPgetVarConflictlengthScore(scip, branchcands[c]);
1050  inferencescore = SCIPgetVarAvgInferenceScore(scip, branchcands[c]);
1051  cutoffscore = SCIPgetVarAvgCutoffScore(scip, branchcands[c]);
1052  gmieffscore = SCIPgetVarAvgGMIScore(scip, branchcands[c]);
1053  lastgmieffscore = SCIPgetVarLastGMIScore(scip, branchcands[c]);
1054  nlscore = calcNlscore(scip, branchruledata->nlcount, branchruledata->nlcountmax, SCIPvarGetProbindex(branchcands[c]));
1055  pscostscore = SCIPgetVarPseudocostScore(scip, branchcands[c], branchcandssol[c]);
1056 
1057  /* replace the pseudo cost score with the already calculated one;
1058  * @todo: use old data for strong branching with propagation?
1059  */
1060  if( SCIPgetVarStrongbranchNode(scip, branchcands[c]) == nodenum )
1061  {
1062  SCIP_Real down;
1063  SCIP_Real up;
1064  SCIP_Real lastlpobjval;
1065  SCIP_Real downgain;
1066  SCIP_Real upgain;
1067 
1068  /* use the score of the strong branching call at the current node */
1069  SCIP_CALL( SCIPgetVarStrongbranchLast(scip, branchcands[c], &down, &up, NULL, NULL, NULL, &lastlpobjval) );
1070  downgain = MAX(down - lastlpobjval, 0.0);
1071  upgain = MAX(up - lastlpobjval, 0.0);
1072  pscostscore = SCIPgetBranchScore(scip, branchcands[c], downgain, upgain);
1073 
1074  SCIPdebugMsg(scip, " -> strong branching on variable <%s> already performed (down=%g (%+g), up=%g (%+g), pscostscore=%g)\n",
1075  SCIPvarGetName(branchcands[c]), down, downgain, up, upgain, pscostscore);
1076  }
1077 
1078  score = calcScore(scip, branchruledata, conflictscore, avgconflictscore, conflengthscore, avgconflengthscore,
1079  inferencescore, avginferencescore, cutoffscore, avgcutoffscore, gmieffscore, lastgmieffscore,
1080  pscostscore, avgpscostscore, nlscore, branchcandsfrac[c], degeneracyfactor);
1081 
1082  /* check for better score of candidate */
1083  if( SCIPisSumGE(scip, score, bestpsscore) )
1084  {
1085  SCIP_Real fracscore;
1086  SCIP_Real domainscore;
1087 
1088  fracscore = MIN(branchcandsfrac[c], 1.0 - branchcandsfrac[c]);
1089  domainscore = -(SCIPvarGetUbLocal(branchcands[c]) - SCIPvarGetLbLocal(branchcands[c]));
1090  if( SCIPisSumGT(scip, score, bestpsscore)
1091  || SCIPisSumGT(scip, fracscore, bestpsfracscore)
1092  || (SCIPisSumGE(scip, fracscore, bestpsfracscore) && domainscore > bestpsdomainscore) )
1093  {
1094  bestpscand = c;
1095  bestpsscore = score;
1096  bestpsfracscore = fracscore;
1097  bestpsdomainscore = domainscore;
1098  }
1099  }
1100  }
1101  }
1102 
1103  for( c = 0; c < nbranchcands; ++c )
1104  {
1105  SCIP_Real conflictscore;
1106  SCIP_Real conflengthscore;
1107  SCIP_Real inferencescore;
1108  SCIP_Real cutoffscore;
1109  SCIP_Real gmieffscore;
1110  SCIP_Real lastgmieffscore;
1111  SCIP_Real pscostscore;
1112  SCIP_Real nlscore;
1113  SCIP_Real score;
1114  SCIP_Bool usesb;
1115  SCIP_Real downgain;
1116  SCIP_Real upgain;
1117  SCIP_Real fracpart;
1118 
1119  assert(branchcands[c] != NULL);
1120  assert(!SCIPisFeasIntegral(scip, branchcandssol[c]));
1121  assert(!SCIPisFeasIntegral(scip, SCIPvarGetLPSol(branchcands[c])));
1122 
1123  /* Record the variables current pseudocosts. These may be overwritten if
1124  * strong branching is performed.
1125  */
1126  sbvars[c] = FALSE;
1127  fracpart = SCIPfeasFrac(scip, SCIPvarGetLPSol(branchcands[c]));
1128  downgain = SCIPgetVarPseudocostVal(scip, branchcands[c], 0.0 - fracpart);
1129  upgain = SCIPgetVarPseudocostVal(scip, branchcands[c], 1.0 - fracpart);
1130  mingains[c] = MIN(downgain, upgain);
1131  maxgains[c] = MAX(downgain, upgain);
1132 
1133  /* get conflict, inference, cutoff, nonlinear, and pseudo cost scores for candidate */
1134  conflictscore = SCIPgetVarConflictScore(scip, branchcands[c]);
1135  conflengthscore = SCIPgetVarConflictlengthScore(scip, branchcands[c]);
1136  inferencescore = SCIPgetVarAvgInferenceScore(scip, branchcands[c]);
1137  cutoffscore = SCIPgetVarAvgCutoffScore(scip, branchcands[c]);
1138  gmieffscore = SCIPgetVarAvgGMIScore(scip, branchcands[c]);
1139  lastgmieffscore = SCIPgetVarLastGMIScore(scip, branchcands[c]);
1140  nlscore = calcNlscore(scip, branchruledata->nlcount, branchruledata->nlcountmax, SCIPvarGetProbindex(branchcands[c]));
1141  pscostscore = SCIPgetVarPseudocostScore(scip, branchcands[c], branchcandssol[c]);
1142  usesb = FALSE;
1143 
1144  /* don't use strong branching on variables that have already been initialized at the current node;
1145  * instead replace the pseudo cost score with the already calculated one;
1146  * @todo: use old data for strong branching with propagation?
1147  */
1148  if( SCIPgetVarStrongbranchNode(scip, branchcands[c]) == nodenum )
1149  {
1150  SCIP_Real down;
1151  SCIP_Real up;
1152  SCIP_Real lastlpobjval;
1153 
1154  /* use the score of the strong branching call at the current node */
1155  SCIP_CALL( SCIPgetVarStrongbranchLast(scip, branchcands[c], &down, &up, NULL, NULL, NULL, &lastlpobjval) );
1156  downgain = MAX(down - lastlpobjval, 0.0);
1157  upgain = MAX(up - lastlpobjval, 0.0);
1158  pscostscore = SCIPgetBranchScore(scip, branchcands[c], downgain, upgain);
1159 
1160  mingains[c] = MIN(downgain, upgain);
1161  maxgains[c] = MAX(downgain, upgain);
1162 
1163  SCIPdebugMsg(scip, " -> strong branching on variable <%s> already performed (down=%g (%+g), up=%g (%+g), pscostscore=%g)\n",
1164  SCIPvarGetName(branchcands[c]), down, downgain, up, upgain, pscostscore);
1165  }
1166  else if( maxninitcands > 0 )
1167  {
1168  SCIP_Real downsize;
1169  SCIP_Real upsize;
1170  SCIP_Real size;
1171 
1172  /* check, if the pseudo cost score of the variable is reliable */
1173  downsize = SCIPgetVarPseudocostCountCurrentRun(scip, branchcands[c], SCIP_BRANCHDIR_DOWNWARDS);
1174  upsize = SCIPgetVarPseudocostCountCurrentRun(scip, branchcands[c], SCIP_BRANCHDIR_UPWARDS);
1175  size = MIN(downsize, upsize);
1176 
1177  /* determine if variable is considered reliable based on the current reliability setting */
1178  /* check fixed number threshold (aka original) reliability first */
1179  assert(!branchruledata->usehyptestforreliability || bestpscand >= 0);
1180  usesb = FALSE;
1181  if( size < reliable )
1182  usesb = TRUE;
1183  else if( branchruledata->userelerrorforreliability && branchruledata->usehyptestforreliability )
1184  {
1185  if( !SCIPisVarPscostRelerrorReliable(scip, branchcands[c], relerrorthreshold, clevel) &&
1186  !SCIPsignificantVarPscostDifference(scip, branchcands[bestpscand], branchcandsfrac[bestpscand],
1187  branchcands[c], branchcandsfrac[c], SCIP_BRANCHDIR_DOWNWARDS, clevel, TRUE) &&
1188  !SCIPsignificantVarPscostDifference(scip, branchcands[bestpscand], 1 - branchcandsfrac[bestpscand],
1189  branchcands[c], 1 - branchcandsfrac[c], SCIP_BRANCHDIR_UPWARDS, clevel, TRUE) )
1190  usesb = TRUE;
1191  }
1192  /* check if relative error is tolerable */
1193  else if( branchruledata->userelerrorforreliability &&
1194  !SCIPisVarPscostRelerrorReliable(scip, branchcands[c], relerrorthreshold, clevel))
1195  usesb = TRUE;
1196  /* check if best pseudo-candidate is significantly better in both directions, use strong-branching otherwise */
1197  else if( branchruledata->usehyptestforreliability &&
1198  !SCIPsignificantVarPscostDifference(scip, branchcands[bestpscand], branchcandsfrac[bestpscand],
1199  branchcands[c], branchcandsfrac[c], SCIP_BRANCHDIR_DOWNWARDS, clevel, TRUE) &&
1200  !SCIPsignificantVarPscostDifference(scip, branchcands[bestpscand], 1 - branchcandsfrac[bestpscand],
1201  branchcands[c], 1 - branchcandsfrac[c], SCIP_BRANCHDIR_UPWARDS, clevel, TRUE))
1202  usesb = TRUE;
1203 
1204  /* count the number of variables that are completely uninitialized */
1205  if( size < 0.1 )
1206  nuninitcands++;
1207  }
1208 
1209  /* combine the five score values */
1210  scoresfrompc[c] = calcScore(scip, branchruledata, 0.0, avgconflictscore, 0.0, avgconflengthscore,
1211  0.0, avginferencescore, 0.0, avgcutoffscore, 0.0, 0.0,
1212  pscostscore, avgpscostscore, 0.0, branchcandsfrac[c], degeneracyfactor);
1213  scoresfromothers[c] = calcScore(scip, branchruledata, conflictscore, avgconflictscore, conflengthscore, avgconflengthscore,
1214  inferencescore, avginferencescore, cutoffscore, avgcutoffscore, gmieffscore, lastgmieffscore,
1215  0.0, avgpscostscore, nlscore, branchcandsfrac[c], degeneracyfactor);
1216  score = scoresfrompc[c] + scoresfromothers[c];
1217  scores[c] = score;
1218  /*score = calcScore(scip, branchruledata, conflictscore, avgconflictscore, conflengthscore, avgconflengthscore,
1219  inferencescore, avginferencescore, cutoffscore, avgcutoffscore, gmieffscore, lastgmieffscore,
1220  pscostscore, avgpscostscore, nlscore, branchcandsfrac[c], degeneracyfactor);*/
1221  if( usesb )
1222  {
1223  int j;
1224 
1225  mingains[c] = 0;
1226  maxgains[c] = 0;
1227  scoresfrompc[c] = 0;
1228  scoresfromothers[c] = 0;
1229 
1230  /* assign a random score to this uninitialized candidate */
1231  if( branchruledata->randinitorder )
1232  score += SCIPrandomGetReal(branchruledata->randnumgen, 0.0, 1e-4);
1233 
1234  /* pseudo cost of variable is not reliable: insert candidate in initcands buffer */
1235  for( j = ninitcands; j > 0 && score > initcandscores[j-1]; --j )
1236  {
1237  initcands[j] = initcands[j-1];
1238  initcandscores[j] = initcandscores[j-1];
1239  }
1240  initcands[j] = c;
1241  initcandscores[j] = score;
1242  ninitcands++;
1243  ninitcands = MIN(ninitcands, maxninitcands);
1244  }
1245  /* in the case of hypothesis reliability, the best pseudo candidate has been determined already */
1246  else if( !branchruledata->usehyptestforreliability )
1247  {
1248  /* variable will keep its pseudo cost value: check for better score of candidate */
1249  if( SCIPisSumGE(scip, score, bestpsscore) )
1250  {
1251  SCIP_Real fracscore;
1252  SCIP_Real domainscore;
1253 
1254  fracscore = MIN(branchcandsfrac[c], 1.0 - branchcandsfrac[c]);
1255  domainscore = -(SCIPvarGetUbLocal(branchcands[c]) - SCIPvarGetLbLocal(branchcands[c]));
1256  if( SCIPisSumGT(scip, score, bestpsscore)
1257  || SCIPisSumGT(scip, fracscore, bestpsfracscore)
1258  || (SCIPisSumGE(scip, fracscore, bestpsfracscore) && domainscore > bestpsdomainscore) )
1259  {
1260  bestpscand = c;
1261  bestpsscore = score;
1262  bestpsfracscore = fracscore;
1263  bestpsdomainscore = domainscore;
1264  }
1265  }
1266  }
1267  }
1268 
1269  /* in the special case that only the best pseudo candidate was selected for strong branching, skip the strong branching */
1270  if( branchruledata->usehyptestforreliability && ninitcands == 1 )
1271  {
1272  ninitcands = 0;
1273  SCIPdebugMsg(scip, "Only one single candidate for initialization-->Skipping strong branching\n");
1274  }
1275 
1276  /* initialize unreliable candidates with strong branching until maxlookahead is reached,
1277  * search best strong branching candidate
1278  */
1279  maxlookahead = (SCIP_Real)branchruledata->maxlookahead * (1.0 + (SCIP_Real)nuninitcands/(SCIP_Real)nbranchcands);
1280  inititer = branchruledata->inititer;
1281  if( inititer == 0 )
1282  {
1283  SCIP_Longint nlpiterations;
1284  SCIP_Longint nlps;
1285 
1286  /* iteration limit is set to twice the average number of iterations spent to resolve a dual feasible SCIP_LP;
1287  * at the first few nodes, this average is not very exact, so we better increase the iteration limit on
1288  * these very important nodes
1289  */
1290  nlpiterations = SCIPgetNDualResolveLPIterations(scip);
1291  nlps = SCIPgetNDualResolveLPs(scip);
1292  if( nlps == 0 )
1293  {
1294  nlpiterations = SCIPgetNNodeInitLPIterations(scip);
1295  nlps = SCIPgetNNodeInitLPs(scip);
1296  if( nlps == 0 )
1297  {
1298  nlpiterations = 1000;
1299  nlps = 1;
1300  }
1301  }
1302  assert(nlps >= 1);
1303  inititer = (int)(2*nlpiterations / nlps);
1304  inititer = (int)((SCIP_Real)inititer * (1.0 + 20.0/nodenum));
1305  inititer = MAX(inititer, 10);
1306  inititer = MIN(inititer, 500);
1307  }
1308 
1309  SCIPdebugMsg(scip, "strong branching (reliable=%g, %d/%d cands, %d uninit, maxcands=%d, maxlookahead=%g, maxbdchgs=%d, inititer=%d, iterations:%" SCIP_LONGINT_FORMAT "/%" SCIP_LONGINT_FORMAT ", basic:%u)\n",
1310  reliable, ninitcands, nbranchcands, nuninitcands, maxninitcands, maxlookahead, maxbdchgs, inititer,
1311  SCIPgetNStrongbranchLPIterations(scip), maxnsblpiterations, SCIPisLPSolBasic(scip));
1312 
1313  bestsbcand = -1;
1314  bestsbscore = -SCIPinfinity(scip);
1315  bestsbfracscore = -SCIPinfinity(scip);
1316  bestsbdomainscore = -SCIPinfinity(scip);
1317  bestuninitsbscore = -SCIPinfinity(scip);
1318  bestuninitsbcand = -1;
1319  lookahead = 0.0;
1320  for( i = 0; i < ninitcands && lookahead < maxlookahead && nbdchgs + nbdconflicts < maxbdchgs
1321  && (i < (int) maxlookahead || SCIPgetNStrongbranchLPIterations(scip) < maxnsblpiterations); ++i )
1322  {
1323  SCIP_Real down;
1324  SCIP_Real up;
1325  SCIP_Real downgain;
1326  SCIP_Real upgain;
1327  SCIP_Bool downvalid;
1328  SCIP_Bool upvalid;
1329  SCIP_Longint ndomredsdown;
1330  SCIP_Longint ndomredsup;
1331  SCIP_Bool lperror;
1332  SCIP_Bool downinf;
1333  SCIP_Bool upinf;
1334  SCIP_Bool downconflict;
1335  SCIP_Bool upconflict;
1336 
1337  /* get candidate number to initialize */
1338  c = initcands[i];
1339  assert(!SCIPisFeasIntegral(scip, branchcandssol[c]));
1340 
1341  if( branchruledata->skipbadinitcands )
1342  {
1343  SCIP_Bool skipsb = FALSE;
1344  /* if the current best candidate is a candidate found by strong branching, determine if candidate pseudo-costs are
1345  * significantly smaller in at least one direction, in which case we safe the execution of strong-branching for now
1346  */
1347  if( bestsbscore > bestpsscore && bestsbscore > bestuninitsbscore && bestsbupvalid && bestsbdownvalid )
1348  {
1349  assert(bestsbcand != -1);
1350  assert(bestsbup != SCIP_INVALID && bestsbdown != SCIP_INVALID); /*lint !e777 lint doesn't like comparing floats */
1351 
1352  /* test if the variable is unlikely to produce a better gain than the currently best one. Skip strong-branching
1353  * in such a case
1354  */
1355  if( SCIPpscostThresholdProbabilityTest(scip, branchcands[c], branchcandsfrac[c], bestsbdown,
1356  SCIP_BRANCHDIR_DOWNWARDS, clevel)
1357  || SCIPpscostThresholdProbabilityTest(scip, branchcands[c], 1.0 - branchcandsfrac[c], bestsbup,
1358  SCIP_BRANCHDIR_UPWARDS, clevel) )
1359  skipsb = TRUE;
1360  }
1361  /* the currently best candidate is also a pseudo-candidate; apply significance test and skip candidate if it
1362  * is significantly worse in at least one direction
1363  */
1364  else if( bestpscand != -1 && bestpsscore > bestuninitsbscore )
1365  {
1366  if( SCIPsignificantVarPscostDifference(scip, branchcands[bestpscand], branchcandsfrac[bestpscand],
1367  branchcands[c], branchcandsfrac[c], SCIP_BRANCHDIR_DOWNWARDS, clevel, TRUE)
1368  || SCIPsignificantVarPscostDifference(scip, branchcands[bestpscand], 1.0 - branchcandsfrac[bestpscand],
1369  branchcands[c], 1.0 - branchcandsfrac[c], SCIP_BRANCHDIR_UPWARDS, clevel, TRUE) )
1370  skipsb = TRUE;
1371  }
1372  /* compare against the best init cand that has been skipped already */
1373  else if( bestuninitsbcand != -1 )
1374  {
1375  if( SCIPsignificantVarPscostDifference(scip, branchcands[bestuninitsbcand], branchcandsfrac[bestuninitsbcand],
1376  branchcands[c], branchcandsfrac[c], SCIP_BRANCHDIR_DOWNWARDS, clevel, TRUE)
1377  || SCIPsignificantVarPscostDifference(scip, branchcands[bestuninitsbcand], 1.0 - branchcandsfrac[bestuninitsbcand],
1378  branchcands[c], 1.0 - branchcandsfrac[c], SCIP_BRANCHDIR_UPWARDS, clevel, TRUE) )
1379  skipsb = TRUE;
1380  }
1381 
1382  /* skip candidate, update the best score of an unitialized candidate */
1383  if( skipsb )
1384  {
1385  if( bestuninitsbcand == -1 )
1386  {
1387  bestuninitsbcand = c;
1388  bestuninitsbscore = initcandscores[i];
1389  }
1390  continue;
1391  }
1392  }
1393  SCIPdebugMsg(scip, "init pseudo cost (%g/%g) of <%s> at %g (score:%g) with strong branching (%d iterations) -- %" SCIP_LONGINT_FORMAT "/%" SCIP_LONGINT_FORMAT " iterations\n",
1396  SCIPvarGetName(branchcands[c]), branchcandssol[c], initcandscores[i],
1397  inititer, SCIPgetNStrongbranchLPIterations(scip), maxnsblpiterations);
1398 
1399  /* use strong branching on candidate */
1400  if( !initstrongbranching )
1401  {
1402  initstrongbranching = TRUE;
1403 
1404  SCIP_CALL( SCIPstartStrongbranch(scip, propagate) );
1405 
1406  /* create arrays for probing-like bound tightening */
1407  if( probingbounds )
1408  {
1409  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &newlbs, nvars) );
1410  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &newubs, nvars) );
1411  }
1412  }
1413 
1414  if( propagate )
1415  {
1416  /* apply strong branching */
1417  SCIP_CALL( SCIPgetVarStrongbranchWithPropagation(scip, branchcands[c], branchcandssol[c], lpobjval, inititer,
1418  branchruledata->maxproprounds, &down, &up, &downvalid, &upvalid, &ndomredsdown, &ndomredsup, &downinf, &upinf,
1419  &downconflict, &upconflict, &lperror, newlbs, newubs) );
1420  }
1421  else
1422  {
1423  /* apply strong branching */
1424  SCIP_CALL( SCIPgetVarStrongbranchFrac(scip, branchcands[c], inititer, FALSE,
1425  &down, &up, &downvalid, &upvalid, &downinf, &upinf, &downconflict, &upconflict, &lperror) );
1426 
1427  ndomredsdown = ndomredsup = 0;
1428  }
1429 
1430  /* check for an error in strong branching */
1431  if( lperror )
1432  {
1433  if( !SCIPisStopped(scip) )
1434  {
1436  "(node %" SCIP_LONGINT_FORMAT ") error in strong branching call for variable <%s> with solution %g\n",
1437  SCIPgetNNodes(scip), SCIPvarGetName(branchcands[c]), branchcandssol[c]);
1438  }
1439  break;
1440  }
1441 
1442  /* Strong branching might have found a new primal solution which updated the cutoff bound. In this case, the
1443  * provedbound computed before can be higher than the cutoffbound and the current node can be cut off.
1444  * Additionally, also if the value for the current best candidate is valid and exceeds the new cutoff bound,
1445  * we want to change the domain of this variable rather than branching on it.
1446  */
1447  if( SCIPgetBestSol(scip) != NULL && SCIPsolGetIndex(SCIPgetBestSol(scip)) != bestsolidx )
1448  {
1449  bestsolidx = SCIPsolGetIndex(SCIPgetBestSol(scip));
1450 
1451  SCIPdebugMsg(scip, " -> strong branching on variable <%s> lead to a new incumbent\n",
1452  SCIPvarGetName(branchcands[c]));
1453 
1454  /* proved bound for current node is larger than new cutoff bound -> cut off current node */
1455  if( SCIPisGE(scip, provedbound, SCIPgetCutoffbound(scip)) )
1456  {
1457  SCIPdebugMsg(scip, " -> node can be cut off (provedbound=%g, cutoff=%g)\n", provedbound, SCIPgetCutoffbound(scip));
1458 
1459  *result = SCIP_CUTOFF;
1460  break; /* terminate initialization loop, because node is infeasible */
1461  }
1462  /* proved bound for down child of best candidate is larger than cutoff bound
1463  * -> increase lower bound of best candidate
1464  * we must only do this if the LP is complete, i.e., we are not doing column generation
1465  */
1466 
1467  else if( bestsbcand != -1 && allcolsinlp )
1468  {
1469  if( !bestsbdowncutoff && bestsbdownvalid && SCIPisGE(scip, bestsbdown, SCIPgetCutoffbound(scip)) )
1470  {
1471  bestsbdowncutoff = TRUE;
1472 
1473  SCIPdebugMsg(scip, " -> valid dual bound for down child of best candidate <%s> is higher than new cutoff bound (valid=%u, bestsbdown=%g, cutoff=%g)\n",
1474  SCIPvarGetName(branchcands[bestsbcand]), bestsbdownvalid, bestsbdown, SCIPgetCutoffbound(scip));
1475 
1476  SCIPdebugMsg(scip, " -> increase lower bound of best candidate <%s> to %g\n",
1477  SCIPvarGetName(branchcands[bestsbcand]), SCIPfeasCeil(scip, branchcandssol[bestsbcand]));
1478 
1479  SCIP_CALL( addBdchg(scip, &bdchginds, &bdchgtypes, &bdchgbounds, &nbdchgs, SCIPvarGetProbindex(branchcands[bestsbcand]),
1480  SCIP_BOUNDTYPE_LOWER, SCIPfeasCeil(scip, branchcandssol[bestsbcand])) );
1481  }
1482  /* proved bound for up child of best candidate is larger than cutoff bound -> decrease upper bound of best candidate */
1483  else if( !bestsbupcutoff && bestsbupvalid && SCIPisGE(scip, bestsbup, SCIPgetCutoffbound(scip)) )
1484  {
1485  bestsbupcutoff = TRUE;
1486 
1487  SCIPdebugMsg(scip, " -> valid dual bound for up child of best candidate <%s> is higher than new cutoff bound (valid=%u, bestsbup=%g, cutoff=%g)\n",
1488  SCIPvarGetName(branchcands[bestsbcand]), bestsbupvalid, bestsbup, SCIPgetCutoffbound(scip));
1489 
1490  SCIPdebugMsg(scip, " -> decrease upper bound of best candidate <%s> to %g\n",
1491  SCIPvarGetName(branchcands[bestsbcand]), SCIPfeasFloor(scip, branchcandssol[bestsbcand]));
1492 
1493  SCIP_CALL( addBdchg(scip, &bdchginds, &bdchgtypes, &bdchgbounds, &nbdchgs, SCIPvarGetProbindex(branchcands[bestsbcand]),
1494  SCIP_BOUNDTYPE_UPPER, SCIPfeasFloor(scip, branchcandssol[bestsbcand])) );
1495  }
1496  }
1497  }
1498 
1499  /* evaluate strong branching */
1500  down = MAX(down, lpobjval);
1501  up = MAX(up, lpobjval);
1502  downgain = down - lpobjval;
1503  upgain = up - lpobjval;
1504  assert(!allcolsinlp || exactsolve || !downvalid || downinf == SCIPisGE(scip, down, SCIPgetCutoffbound(scip)));
1505  assert(!allcolsinlp || exactsolve || !upvalid || upinf == SCIPisGE(scip, up, SCIPgetCutoffbound(scip)));
1506  assert(downinf || !downconflict);
1507  assert(upinf || !upconflict);
1508 
1509  /* @todo: store pseudo cost only for valid bounds?
1510  * depending on the user parameter choice of storesemiinitcosts, pseudo costs are also updated in single directions,
1511  * if the node in the other direction was infeasible or cut off
1512  */
1513  if( !downinf
1514 #ifdef WITH_LPSOLSTAT
1516 #endif
1517  && (!upinf || branchruledata->storesemiinitcosts) )
1518  {
1519  SCIP_Real weight;
1520 
1521  /* smaller weights are given if the strong branching hit the time limit in the corresponding direction */
1522  if( branchruledata->usesmallweightsitlim )
1524  else
1525  weight = 1.0;
1526 
1527  /* update pseudo cost values */
1528  SCIP_CALL( SCIPupdateVarPseudocostSymmetric(scip, branchruledata, branchcands[c], branchorbitidx, c, 0.0 - branchcandsfrac[c], downgain, weight) );
1529  }
1530  if( !upinf
1531 #ifdef WITH_LPSOLSTAT
1533 #endif
1534  && (!downinf || branchruledata->storesemiinitcosts) )
1535  {
1536  SCIP_Real weight;
1537 
1538  /* smaller weights are given if the strong branching hit the time limit in the corresponding direction */
1539  if( branchruledata->usesmallweightsitlim )
1541  else
1542  weight = 1.0;
1543 
1544  /* update pseudo cost values */
1545  SCIP_CALL( SCIPupdateVarPseudocostSymmetric(scip, branchruledata, branchcands[c], branchorbitidx, c, 1.0 - branchcandsfrac[c], upgain, weight) );
1546  }
1547 
1548  /* the minimal lower bound of both children is a proved lower bound of the current subtree */
1549  if( allcolsinlp && !exactsolve && downvalid && upvalid )
1550  {
1551  SCIP_Real minbound;
1552 
1553  minbound = MIN(down, up);
1554  provedbound = MAX(provedbound, minbound);
1555  assert((downinf && upinf) || SCIPisLT(scip, provedbound, SCIPgetCutoffbound(scip)));
1556 
1557  /* save probing-like bounds detected during strong branching */
1558  if( probingbounds )
1559  {
1560  int v;
1561 
1562  assert(newlbs != NULL);
1563  assert(newubs != NULL);
1564 
1565  for( v = 0; v < nvars; ++v )
1566  {
1567  if( SCIPisGT(scip, newlbs[v], SCIPvarGetLbLocal(vars[v])) )
1568  {
1569  SCIPdebugMsg(scip, "better lower bound for variable <%s>: %.9g -> %.9g (by strongbranching on <%s>)\n",
1570  SCIPvarGetName(vars[v]), SCIPvarGetLbLocal(vars[v]), newlbs[v], SCIPvarGetName(branchcands[c]));
1571 
1572  SCIP_CALL( addBdchg(scip, &bdchginds, &bdchgtypes, &bdchgbounds, &nbdchgs, v,
1573  SCIP_BOUNDTYPE_LOWER, newlbs[v]) );
1574  }
1575  if( SCIPisLT(scip, newubs[v], SCIPvarGetUbLocal(vars[v])) )
1576  {
1577  SCIPdebugMsg(scip, "better upper bound for variable <%s>: %.9g -> %.9g (by strongbranching on <%s>)\n",
1578  SCIPvarGetName(vars[v]), SCIPvarGetUbLocal(vars[v]), newubs[v], SCIPvarGetName(branchcands[c]));
1579 
1580  SCIP_CALL( addBdchg(scip, &bdchginds, &bdchgtypes, &bdchgbounds, &nbdchgs, v,
1581  SCIP_BOUNDTYPE_UPPER, newubs[v]) );
1582  }
1583  }
1584  }
1585  }
1586 
1587  /* check if there are infeasible roundings */
1588  if( downinf || upinf )
1589  {
1590  assert(allcolsinlp || propagate);
1591  assert(!exactsolve);
1592 
1593  if( downinf && upinf )
1594  {
1595  /* both roundings are infeasible -> node is infeasible */
1596  SCIPdebugMsg(scip, " -> variable <%s> is infeasible in both directions (conflict: %u/%u)\n",
1597  SCIPvarGetName(branchcands[c]), downconflict, upconflict);
1598  *result = SCIP_CUTOFF;
1599  break; /* terminate initialization loop, because node is infeasible */
1600  }
1601  else
1602  {
1603  /* rounding is infeasible in one direction -> round variable in other direction */
1604  SCIPdebugMsg(scip, " -> variable <%s> is infeasible in %s branch (conflict: %u/%u)\n",
1605  SCIPvarGetName(branchcands[c]), downinf ? "downward" : "upward", downconflict, upconflict);
1606  SCIP_CALL( addBdchg(scip, &bdchginds, &bdchgtypes, &bdchgbounds, &nbdchgs, SCIPvarGetProbindex(branchcands[c]),
1608  (downinf ? SCIPfeasCeil(scip, branchcandssol[c]) : SCIPfeasFloor(scip, branchcandssol[c]))) );
1609  if( nbdchgs + nbdconflicts >= maxbdchgs )
1610  break; /* terminate initialization loop, because enough roundings are performed */
1611  }
1612  }
1613  else
1614  {
1615  SCIP_Real conflictscore;
1616  SCIP_Real conflengthscore;
1617  SCIP_Real inferencescore;
1618  SCIP_Real cutoffscore;
1619  SCIP_Real gmieffscore;
1620  SCIP_Real lastgmieffscore;
1621  SCIP_Real pscostscore;
1622  SCIP_Real nlscore;
1623  SCIP_Real score;
1624 
1625  mingains[c] = MIN(downgain, upgain);
1626  maxgains[c] = MAX(downgain, upgain);
1627 
1628  sbdown[c] = down;
1629  sbup[c] = up;
1630  sbdownvalid[c] = downvalid;
1631  sbupvalid[c] = upvalid;
1632  sbvars[c] = TRUE;
1633 
1634  /* check for a better score */
1635  conflictscore = SCIPgetVarConflictScore(scip, branchcands[c]);
1636  conflengthscore = SCIPgetVarConflictlengthScore(scip, branchcands[c]);
1637  nlscore = calcNlscore(scip, branchruledata->nlcount, branchruledata->nlcountmax, SCIPvarGetProbindex(branchcands[c]));
1638 
1639  /* optionally, use only local information obtained via strong branching for this candidate, i.e., local
1640  * domain reductions and no cutoff score
1641  */
1642  inferencescore = branchruledata->usesblocalinfo ? SCIPgetBranchScore(scip, branchcands[c], (SCIP_Real)ndomredsdown, (SCIP_Real)ndomredsup)
1643  : SCIPgetVarAvgInferenceScore(scip, branchcands[c]);
1644  cutoffscore = branchruledata->usesblocalinfo ? 0.0 : SCIPgetVarAvgCutoffScore(scip, branchcands[c]);
1645  gmieffscore = branchruledata->usesblocalinfo ? 0.0 : SCIPgetVarAvgGMIScore(scip, branchcands[c]);
1646  lastgmieffscore = branchruledata->usesblocalinfo ? 0.0 : SCIPgetVarLastGMIScore(scip, branchcands[c]);
1647  pscostscore = SCIPgetBranchScore(scip, branchcands[c], downgain, upgain);
1648 
1649  scoresfrompc[c] = calcScore(scip, branchruledata, 0.0, avgconflictscore, 0.0, avgconflengthscore,
1650  0.0, avginferencescore, 0.0, avgcutoffscore, 0.0, 0.0, pscostscore,
1651  avgpscostscore, 0.0, branchcandsfrac[c], degeneracyfactor);
1652  scoresfromothers[c] = calcScore(scip, branchruledata, conflictscore, avgconflictscore, conflengthscore, avgconflengthscore,
1653  inferencescore, avginferencescore, cutoffscore, avgcutoffscore, gmieffscore,
1654  lastgmieffscore, 0.0, avgpscostscore, nlscore, branchcandsfrac[c],
1655  degeneracyfactor);
1656  score = scoresfrompc[c] + scoresfromothers[c];
1657  scores[c] = score;
1658  /*score = calcScore(scip, branchruledata, conflictscore, avgconflictscore, conflengthscore, avgconflengthscore,
1659  inferencescore, avginferencescore, cutoffscore, avgcutoffscore, gmieffscore, lastgmieffscore,
1660  pscostscore, avgpscostscore, nlscore, branchcandsfrac[c], degeneracyfactor);*/
1661 
1662  if( SCIPisSumGE(scip, score, bestsbscore) )
1663  {
1664  SCIP_Real fracscore;
1665  SCIP_Real domainscore;
1666 
1667  fracscore = MIN(branchcandsfrac[c], 1.0 - branchcandsfrac[c]);
1668  domainscore = -(SCIPvarGetUbLocal(branchcands[c]) - SCIPvarGetLbLocal(branchcands[c]));
1669  if( SCIPisSumGT(scip, score, bestsbscore)
1670  || SCIPisSumGT(scip, fracscore, bestsbfracscore)
1671  || (SCIPisSumGE(scip, fracscore, bestsbfracscore) && domainscore > bestsbdomainscore) )
1672  {
1673  bestsbcand = c;
1674  bestsbscore = score;
1675  bestsbdown = down;
1676  bestsbup = up;
1677  bestsbdownvalid = downvalid;
1678  bestsbupvalid = upvalid;
1679  bestsbdowncutoff = FALSE;
1680  bestsbupcutoff = FALSE;
1681  bestsbfracscore = fracscore;
1682  bestsbdomainscore = domainscore;
1683  lookahead = 0.0;
1684  }
1685  else
1686  lookahead += 0.5;
1687  }
1688  else
1689  lookahead += 1.0;
1690 
1691  SCIPdebugMsg(scip, " -> variable <%s> (solval=%g, down=%g (%+g,valid=%u), up=%g (%+g,valid=%u), score=%g/ %g/%g %g/%g %g -> %g)\n",
1692  SCIPvarGetName(branchcands[c]), branchcandssol[c], down, downgain, downvalid, up, upgain, upvalid,
1693  pscostscore, conflictscore, conflengthscore, inferencescore, cutoffscore, gmieffscore, score);
1694  }
1695  }
1696 #ifdef SCIP_DEBUG
1697  if( bestsbcand >= 0 )
1698  {
1699  SCIPdebugMsg(scip, " -> best: <%s> (%g / %g / %g), lookahead=%g/%g\n",
1700  SCIPvarGetName(branchcands[bestsbcand]), bestsbscore, bestsbfracscore, bestsbdomainscore,
1701  lookahead, maxlookahead);
1702  }
1703 #endif
1704 
1705  if( initstrongbranching )
1706  {
1707  if( probingbounds )
1708  {
1709  assert(newlbs != NULL);
1710  assert(newubs != NULL);
1711 
1712  SCIPfreeBlockMemoryArray(scip, &newubs, nvars);
1713  SCIPfreeBlockMemoryArray(scip, &newlbs, nvars);
1714  }
1715 
1716  SCIP_CALL( SCIPendStrongbranch(scip) );
1717 
1719  {
1720  assert(SCIPhasCurrentNodeLP(scip));
1721  *result = SCIP_CUTOFF;
1722  }
1723  }
1724 
1725  if( *result != SCIP_CUTOFF )
1726  {
1727  /* get the score of the best uninitialized strong branching candidate */
1728  if( i < ninitcands && bestuninitsbcand == -1 )
1729  bestuninitsbscore = initcandscores[i];
1730 
1731  /* if the best pseudo cost candidate is better than the best uninitialized strong branching candidate,
1732  * compare it to the best initialized strong branching candidate
1733  */
1734  if( bestpsscore > bestuninitsbscore && SCIPisSumGT(scip, bestpsscore, bestsbscore) )
1735  {
1736  bestcand = bestpscand;
1737  bestisstrongbranch = FALSE;
1738  }
1739  else if( bestsbcand >= 0 )
1740  {
1741  bestcand = bestsbcand;
1742  bestisstrongbranch = TRUE;
1743  }
1744  else
1745  {
1746  /* no candidate was initialized, and the best score is the one of the first candidate in the initialization
1747  * queue
1748  */
1749  assert(ninitcands >= 1);
1750  bestcand = initcands[0];
1751  bestisstrongbranch = FALSE;
1752  }
1753 
1754  /* update lower bound of current node */
1755  if( allcolsinlp && !exactsolve )
1756  {
1757  assert(SCIPisLT(scip, provedbound, SCIPgetCutoffbound(scip)));
1758  SCIP_CALL( SCIPupdateNodeLowerbound(scip, SCIPgetCurrentNode(scip), provedbound) );
1759  }
1760  }
1761 
1762  /* apply domain reductions */
1763  if( nbdchgs > 0 )
1764  {
1765  if( *result != SCIP_CUTOFF )
1766  {
1767  SCIP_CALL( applyBdchgs(scip, vars, bdchginds, bdchgtypes, bdchgbounds, nbdchgs, result) );
1768  if( *result != SCIP_CUTOFF )
1769  *result = SCIP_REDUCEDDOM;
1770  }
1771  freeBdchgs(scip, &bdchginds, &bdchgtypes, &bdchgbounds, &nbdchgs);
1772  }
1773 
1774  /* Apply the Treemodel branching rule to potentially select a better branching candidate than the current one. */
1775  if( *result != SCIP_CUTOFF && *result != SCIP_REDUCEDDOM && *result != SCIP_CONSADDED && SCIPtreemodelIsEnabled(scip, branchruledata->treemodel) )
1776  {
1777  SCIP_Real smallpscost;
1778  SCIP_Bool usetreemodel;
1779 
1780  usetreemodel = TRUE;
1781 
1782  /* If the pseudocosts are zero, use SCIPs best variable since the Treemodel is not applicable */
1783  if( SCIPisZero(scip, maxgains[bestcand]))
1784  {
1785  usetreemodel = FALSE;
1786  }
1787 
1788  /* If SCIPs best candidate was selected due to hybrid branching scores
1789  * rather than because of pseudocosts, then we keep it.
1790  */
1791  SCIP_CALL( SCIPgetRealParam(scip, "branching/treemodel/smallpscost", &smallpscost) );
1792  if( usetreemodel == TRUE && avgpscostscore <= smallpscost )
1793  {
1794  int cand;
1795  for( cand = 0; cand < nbranchcands; ++cand )
1796  {
1797  if( scoresfrompc[cand] > scoresfrompc[bestcand] )
1798  {
1799  usetreemodel = FALSE;
1800  break;
1801  }
1802  }
1803  }
1804 
1805  if( usetreemodel == TRUE )
1806  {
1808  scip, /* SCIP data structure */
1809  branchruledata->treemodel, /* branching rule */
1810  branchcands, /* branching candidate storage */
1811  mingains, /* minimum gain of rounding downwards or upwards */
1812  maxgains, /* maximum gain of rounding downwards or upwards */
1813  scoresfromothers, /* scores from other branching methods */
1814  nbranchcands, /* the number of branching candidates */
1815  &bestcand /* the best branching candidate found by SCIP */
1816  ) );
1817  }
1818  }
1819 
1820  /* free buffer for the lp gains and pseudocost scores */
1821  SCIPfreeBufferArray(scip, &scoresfromothers);
1822  SCIPfreeBufferArray(scip, &scoresfrompc);
1823  SCIPfreeBufferArray(scip, &scores);
1824  SCIPfreeBufferArray(scip, &maxgains);
1825  SCIPfreeBufferArray(scip, &mingains);
1826 
1827  /* free buffer for the unreliable candidates */
1828  SCIPfreeBufferArray(scip, &initcandscores);
1829  SCIPfreeBufferArray(scip, &initcands);
1830  }
1831 
1832  /* if no domain could be reduced, create the branching */
1833  if( *result != SCIP_CUTOFF && *result != SCIP_REDUCEDDOM && *result != SCIP_CONSADDED && executebranch )
1834  {
1835  SCIP_NODE* downchild;
1836  SCIP_NODE* upchild;
1837  SCIP_VAR* var;
1838  SCIP_Real val;
1839 
1840  assert(*result == SCIP_DIDNOTRUN);
1841  assert(0 <= bestcand && bestcand < nbranchcands);
1842  assert(!SCIPisFeasIntegral(scip, branchcandssol[bestcand]));
1843  assert(!allcolsinlp || SCIPisLT(scip, provedbound, SCIPgetCutoffbound(scip)));
1844  assert(!bestsbdowncutoff && !bestsbupcutoff);
1845 
1846  var = branchcands[bestcand];
1847  val = branchcandssol[bestcand];
1848 
1849  /* perform the branching */
1850  SCIPdebugMsg(scip, " -> %d (%d) cands, sel cand %d: var <%s> (sol=%g, down=%g (%+g), up=%g (%+g), sb=%u, psc=%g/%g [%g])\n",
1851  nbranchcands, ninitcands, bestcand, SCIPvarGetName(var), branchcandssol[bestcand],
1852  bestsbdown, bestsbdown - lpobjval, bestsbup, bestsbup - lpobjval, bestisstrongbranch,
1855  SCIPgetVarPseudocostScoreCurrentRun(scip, var, branchcandssol[bestcand]));
1856  SCIP_UNUSED(bestisstrongbranch);
1857  SCIP_CALL( SCIPbranchVarVal(scip, var, val, &downchild, NULL, &upchild) );
1858  assert(downchild != NULL);
1859  assert(upchild != NULL);
1860 
1861  /* update the lower bounds in the children */
1862  if( sbvars[bestcand] && allcolsinlp && !exactsolve )
1863  {
1864  if( sbdownvalid[bestcand] )
1865  {
1866  assert(SCIPisLT(scip, sbdown[bestcand], SCIPgetCutoffbound(scip)));
1867  SCIP_CALL( SCIPupdateNodeLowerbound(scip, downchild, sbdown[bestcand]) );
1868  assert(SCIPisGE(scip, SCIPgetNodeLowerbound(scip, downchild), provedbound));
1869  }
1870  if( sbupvalid[bestcand] )
1871  {
1872  assert(SCIPisLT(scip, sbup[bestcand], SCIPgetCutoffbound(scip)));
1873  SCIP_CALL( SCIPupdateNodeLowerbound(scip, upchild, sbup[bestcand]) );
1874  assert(SCIPisGE(scip, SCIPgetNodeLowerbound(scip, upchild), provedbound));
1875  }
1876  }
1877 
1878  SCIPdebugMsg(scip, " -> down child's lowerbound: %g\n", SCIPnodeGetLowerbound(downchild));
1879  SCIPdebugMsg(scip, " -> up child's lowerbound : %g\n", SCIPnodeGetLowerbound(upchild));
1880 
1882 
1883  *result = SCIP_BRANCHED;
1884  }
1885 
1886  /* free buffer for the strong branching lp gains */
1887  SCIPfreeBufferArray(scip, &sbvars);
1888  SCIPfreeBufferArray(scip, &sbupvalid);
1889  SCIPfreeBufferArray(scip, &sbdownvalid);
1890  SCIPfreeBufferArray(scip, &sbup);
1891  SCIPfreeBufferArray(scip, &sbdown);
1892 
1893  return SCIP_OKAY;
1894 }
1895 
1896 
1897 /*
1898  * Callback methods
1899  */
1900 
1901 /** copy method for branchrule plugins (called when SCIP copies plugins) */
1902 static
1903 SCIP_DECL_BRANCHCOPY(branchCopyRelpscost)
1904 { /*lint --e{715}*/
1905  assert(scip != NULL);
1906  assert(branchrule != NULL);
1907  assert(strcmp(SCIPbranchruleGetName(branchrule), BRANCHRULE_NAME) == 0);
1908 
1909  /* call inclusion method of branchrule */
1911 
1912  return SCIP_OKAY;
1913 }
1914 
1915 /** destructor of branching rule to free user data (called when SCIP is exiting) */
1916 static
1917 SCIP_DECL_BRANCHFREE(branchFreeRelpscost)
1918 { /*lint --e{715}*/
1919  SCIP_BRANCHRULEDATA* branchruledata;
1920  branchruledata = SCIPbranchruleGetData(branchrule);
1922  /* free Treemodel parameter data structure */
1923  SCIP_CALL( SCIPtreemodelFree(scip, &branchruledata->treemodel) );
1924 
1925  /* free branching rule data */
1926  SCIPfreeBlockMemory(scip, &branchruledata);
1927  SCIPbranchruleSetData(branchrule, NULL);
1928 
1929  return SCIP_OKAY;
1930 }
1931 
1932 
1933 /** solving process initialization method of branching rule (called when branch and bound process is about to begin) */
1934 static
1935 SCIP_DECL_BRANCHINITSOL(branchInitsolRelpscost)
1936 { /*lint --e{715}*/
1937  SCIP_BRANCHRULEDATA* branchruledata;
1938 
1939  /* initialize branching rule data */
1940  branchruledata = SCIPbranchruleGetData(branchrule);
1941  branchruledata->nlcount = NULL;
1942  branchruledata->nlcountsize = 0;
1943  branchruledata->nlcountmax = 1;
1944  assert(branchruledata->startrandseed >= 0);
1945 
1946  /* create a random number generator */
1947  SCIP_CALL( SCIPcreateRandom(scip, &branchruledata->randnumgen,
1948  (unsigned int)branchruledata->startrandseed, TRUE) );
1949 
1950  return SCIP_OKAY;
1951 }
1952 
1953 
1954 /** solving process deinitialization method of branching rule (called before branch and bound process data is freed) */
1955 static
1956 SCIP_DECL_BRANCHEXITSOL(branchExitsolRelpscost)
1957 { /*lint --e{715}*/
1958  SCIP_BRANCHRULEDATA* branchruledata;
1959 
1960  /* free memory in branching rule data */
1961  branchruledata = SCIPbranchruleGetData(branchrule);
1962  SCIPfreeBlockMemoryArrayNull(scip, &branchruledata->nlcount, branchruledata->nlcountsize);
1963 
1964  /* free random number generator */
1965  SCIPfreeRandom(scip, &branchruledata->randnumgen);
1966 
1967  SCIPfreeBlockMemoryArrayNull(scip, &branchruledata->orbitrep, branchruledata->npermvars);
1968  SCIPfreeBlockMemoryArrayNull(scip, &branchruledata->varorbitmap, branchruledata->npermvars);
1969  SCIPfreeBlockMemoryArrayNull(scip, &branchruledata->orbitbegins, branchruledata->npermvars);
1970  SCIPfreeBlockMemoryArrayNull(scip, &branchruledata->orbits, branchruledata->npermvars);
1971  branchruledata->nosymmetry = FALSE;
1972  branchruledata->norbits = 0;
1973  branchruledata->permvars = NULL;
1974  branchruledata->permvarmap = NULL;
1975  branchruledata->npermvars = 0;
1976 
1977  return SCIP_OKAY;
1978 }
1979 
1980 
1981 /** branching execution method for fractional LP solutions */
1982 static
1983 SCIP_DECL_BRANCHEXECLP(branchExeclpRelpscost)
1984 { /*lint --e{715}*/
1985  SCIP_BRANCHRULEDATA* branchruledata;
1986  SCIP_VAR** lpcands;
1987  SCIP_Real* lpcandssol;
1988  SCIP_Real* lpcandsfrac;
1989  SCIP_VAR** filteredlpcands;
1990  SCIP_Real* filteredlpcandssol;
1991  SCIP_Real* filteredlpcandsfrac;
1992  SCIP_Bool runfiltering;
1993  int* filteredlpcandsorbitidx = NULL;
1994  int nfilteredlpcands;
1995  int nlpcands;
1996 
1997  assert(branchrule != NULL);
1998  assert(strcmp(SCIPbranchruleGetName(branchrule), BRANCHRULE_NAME) == 0);
1999  assert(scip != NULL);
2000  assert(result != NULL);
2001 
2002  SCIPdebugMsg(scip, "Execlp method of relpscost branching in node %" SCIP_LONGINT_FORMAT "\n", SCIPnodeGetNumber(SCIPgetCurrentNode(scip)));
2003 
2005  {
2006  *result = SCIP_DIDNOTRUN;
2007  SCIPdebugMsg(scip, "Could not apply relpscost branching, as the current LP was not solved to optimality.\n");
2008 
2009  return SCIP_OKAY;
2010  }
2011 
2012  /* get branching candidates */
2013  SCIP_CALL( SCIPgetLPBranchCands(scip, &lpcands, &lpcandssol, &lpcandsfrac, NULL, &nlpcands, NULL) );
2014  assert(nlpcands > 0);
2015 
2016  branchruledata = SCIPbranchruleGetData(branchrule);
2017  assert(branchruledata != NULL);
2018 
2019  /* determine whether we should run filtering */
2020  runfiltering = ! branchruledata->nosymmetry && branchruledata->filtercandssym && SCIPgetSubscipDepth(scip) == 0 && ! SCIPinDive(scip) && ! SCIPinProbing(scip);
2021 
2022  /* init orbits if necessary */
2023  if( runfiltering )
2024  {
2025  SCIP_CALL( initOrbits(scip, branchruledata) );
2026  }
2027 
2028  /* determine fractional variables (possibly filter by using symmetries) */
2029  if( runfiltering && branchruledata->norbits != 0 )
2030  {
2031  SCIP_CALL( SCIPallocBufferArray(scip, &filteredlpcands, nlpcands) );
2032  SCIP_CALL( SCIPallocBufferArray(scip, &filteredlpcandssol, nlpcands) );
2033  SCIP_CALL( SCIPallocBufferArray(scip, &filteredlpcandsfrac, nlpcands) );
2034  SCIP_CALL( SCIPallocBufferArray(scip, &filteredlpcandsorbitidx, nlpcands) );
2035 
2036  /* determine filtered fractional variables */
2037  SCIP_CALL( filterSymmetricVariables(scip, branchruledata, lpcands, lpcandssol, lpcandsfrac, nlpcands,
2038  filteredlpcands, filteredlpcandssol, filteredlpcandsfrac, filteredlpcandsorbitidx, &nfilteredlpcands) );
2039  }
2040  else
2041  {
2042  /* No orbits available. Copy all (unfiltered) branching candidates, because they will be updated w.r.t. the strong branching LP solution */
2043  SCIP_CALL( SCIPduplicateBufferArray(scip, &filteredlpcands, lpcands, nlpcands) );
2044  SCIP_CALL( SCIPduplicateBufferArray(scip, &filteredlpcandssol, lpcandssol, nlpcands) );
2045  SCIP_CALL( SCIPduplicateBufferArray(scip, &filteredlpcandsfrac, lpcandsfrac, nlpcands) );
2046  nfilteredlpcands = nlpcands;
2047  }
2048 
2049  /* execute branching rule */
2050  SCIP_CALL( execRelpscost(scip, branchrule, filteredlpcands, filteredlpcandssol, filteredlpcandsfrac, filteredlpcandsorbitidx, nfilteredlpcands, TRUE, result) );
2051 
2052  SCIPfreeBufferArrayNull(scip, &filteredlpcandsorbitidx);
2053  SCIPfreeBufferArray(scip, &filteredlpcandsfrac);
2054  SCIPfreeBufferArray(scip, &filteredlpcandssol);
2055  SCIPfreeBufferArray(scip, &filteredlpcands);
2056 
2057  return SCIP_OKAY;
2058 }
2059 
2060 
2061 /*
2062  * branching specific interface methods
2063  */
2064 
2065 /** creates the reliable pseudo cost branching rule and includes it in SCIP */
2067  SCIP* scip /**< SCIP data structure */
2068  )
2069 {
2070  SCIP_BRANCHRULEDATA* branchruledata;
2071  SCIP_BRANCHRULE* branchrule;
2072 
2073  /* create relpscost branching rule data */
2074  SCIP_CALL( SCIPallocBlockMemory(scip, &branchruledata) );
2075 
2076  branchruledata->nosymmetry = FALSE;
2077  branchruledata->orbits = NULL;
2078  branchruledata->orbitbegins = NULL;
2079  branchruledata->orbitrep = NULL;
2080  branchruledata->varorbitmap = NULL;
2081  branchruledata->norbits = 0;
2082  branchruledata->permvars = NULL;
2083  branchruledata->npermvars = 0;
2084  branchruledata->permvarmap = NULL;
2085 
2086  /* include branching rule */
2088  BRANCHRULE_MAXDEPTH, BRANCHRULE_MAXBOUNDDIST, branchruledata) );
2089 
2090  assert(branchrule != NULL);
2091 
2092  /* set non-fundamental callbacks via specific setter functions*/
2093  SCIP_CALL( SCIPsetBranchruleCopy(scip, branchrule, branchCopyRelpscost) );
2094  SCIP_CALL( SCIPsetBranchruleFree(scip, branchrule, branchFreeRelpscost) );
2095  SCIP_CALL( SCIPsetBranchruleInitsol(scip, branchrule, branchInitsolRelpscost) );
2096  SCIP_CALL( SCIPsetBranchruleExitsol(scip, branchrule, branchExitsolRelpscost) );
2097  SCIP_CALL( SCIPsetBranchruleExecLp(scip, branchrule, branchExeclpRelpscost) );
2098 
2099  /* relpscost branching rule parameters */
2101  "branching/relpscost/conflictweight",
2102  "weight in score calculations for conflict score",
2103  &branchruledata->conflictweight, TRUE, DEFAULT_CONFLICTWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2105  "branching/relpscost/conflictlengthweight",
2106  "weight in score calculations for conflict length score",
2107  &branchruledata->conflengthweight, TRUE, DEFAULT_CONFLENGTHWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2109  "branching/relpscost/inferenceweight",
2110  "weight in score calculations for inference score",
2111  &branchruledata->inferenceweight, TRUE, DEFAULT_INFERENCEWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2113  "branching/relpscost/cutoffweight",
2114  "weight in score calculations for cutoff score",
2115  &branchruledata->cutoffweight, TRUE, DEFAULT_CUTOFFWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2117  "branching/relpscost/gmiavgeffweight",
2118  "weight in score calculations for average GMI cuts normalized efficacy",
2119  &branchruledata->gmiavgeffweight, TRUE, DEFAULT_GMIAVGEFFWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2121  "branching/relpscost/gmilasteffweight",
2122  "weight in score calculations for last GMI cuts normalized efficacy",
2123  &branchruledata->gmilasteffweight, TRUE, DEFAULT_GMILASTEFFWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2125  "branching/relpscost/pscostweight",
2126  "weight in score calculations for pseudo cost score",
2127  &branchruledata->pscostweight, TRUE, DEFAULT_PSCOSTWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2129  "branching/relpscost/nlscoreweight",
2130  "weight in score calculations for nlcount score",
2131  &branchruledata->nlscoreweight, TRUE, DEFAULT_NLSCOREWEIGHT, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
2133  "branching/relpscost/minreliable",
2134  "minimal value for minimum pseudo cost size to regard pseudo cost value as reliable",
2135  &branchruledata->minreliable, TRUE, DEFAULT_MINRELIABLE, 0.0, SCIP_REAL_MAX, NULL, NULL) );
2137  "branching/relpscost/maxreliable",
2138  "maximal value for minimum pseudo cost size to regard pseudo cost value as reliable",
2139  &branchruledata->maxreliable, TRUE, DEFAULT_MAXRELIABLE, 0.0, SCIP_REAL_MAX, NULL, NULL) );
2141  "branching/relpscost/sbiterquot",
2142  "maximal fraction of strong branching LP iterations compared to node relaxation LP iterations",
2143  &branchruledata->sbiterquot, FALSE, DEFAULT_SBITERQUOT, 0.0, SCIP_REAL_MAX, NULL, NULL) );
2144  SCIP_CALL( SCIPaddIntParam(scip,
2145  "branching/relpscost/sbiterofs",
2146  "additional number of allowed strong branching LP iterations",
2147  &branchruledata->sbiterofs, FALSE, DEFAULT_SBITEROFS, 0, INT_MAX, NULL, NULL) );
2148  SCIP_CALL( SCIPaddIntParam(scip,
2149  "branching/relpscost/maxlookahead",
2150  "maximal number of further variables evaluated without better score",
2151  &branchruledata->maxlookahead, TRUE, DEFAULT_MAXLOOKAHEAD, 1, INT_MAX, NULL, NULL) );
2152  SCIP_CALL( SCIPaddIntParam(scip,
2153  "branching/relpscost/initcand",
2154  "maximal number of candidates initialized with strong branching per node",
2155  &branchruledata->initcand, FALSE, DEFAULT_INITCAND, 0, INT_MAX, NULL, NULL) );
2156  SCIP_CALL( SCIPaddIntParam(scip,
2157  "branching/relpscost/inititer",
2158  "iteration limit for strong branching initializations of pseudo cost entries (0: auto)",
2159  &branchruledata->inititer, FALSE, DEFAULT_INITITER, 0, INT_MAX, NULL, NULL) );
2160  SCIP_CALL( SCIPaddIntParam(scip,
2161  "branching/relpscost/maxbdchgs",
2162  "maximal number of bound tightenings before the node is reevaluated (-1: unlimited)",
2163  &branchruledata->maxbdchgs, TRUE, DEFAULT_MAXBDCHGS, -1, INT_MAX, NULL, NULL) );
2164  SCIP_CALL( SCIPaddIntParam(scip,
2165  "branching/relpscost/maxproprounds",
2166  "maximum number of propagation rounds to be performed during strong branching before solving the LP (-1: no limit, -2: parameter settings)",
2167  &branchruledata->maxproprounds, TRUE, DEFAULT_MAXPROPROUNDS, -2, INT_MAX, NULL, NULL) );
2169  "branching/relpscost/probingbounds",
2170  "should valid bounds be identified in a probing-like fashion during strong branching (only with propagation)?",
2171  &branchruledata->probingbounds, TRUE, DEFAULT_PROBINGBOUNDS, NULL, NULL) );
2172 
2173  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/userelerrorreliability",
2174  "should reliability be based on relative errors?", &branchruledata->userelerrorforreliability, TRUE, DEFAULT_USERELERRORFORRELIABILITY,
2175  NULL, NULL) );
2176 
2177  SCIP_CALL( SCIPaddRealParam(scip, "branching/relpscost/lowerrortol", "low relative error tolerance for reliability",
2178  &branchruledata->lowerrortol, TRUE, DEFAULT_LOWERRORTOL, 0.0, SCIP_REAL_MAX, NULL, NULL) );
2179 
2180  SCIP_CALL( SCIPaddRealParam(scip, "branching/relpscost/higherrortol", "high relative error tolerance for reliability",
2181  &branchruledata->higherrortol, TRUE, DEFAULT_HIGHERRORTOL, 0.0, SCIP_REAL_MAX, NULL, NULL) );
2182 
2183 /**! [SnippetCodeStyleParenIndent] */
2184 
2185  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/storesemiinitcosts",
2186  "should strong branching result be considered for pseudo costs if the other direction was infeasible?",
2187  &branchruledata->storesemiinitcosts, TRUE, DEFAULT_STORESEMIINITCOSTS,
2188  NULL, NULL) );
2189 
2190 /**! [SnippetCodeStyleParenIndent] */
2191 
2192  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/usesblocalinfo",
2193  "should the scoring function use only local cutoff and inference information obtained for strong branching candidates?",
2194  &branchruledata->usesblocalinfo, TRUE, DEFAULT_USESBLOCALINFO,
2195  NULL, NULL) );
2196 
2197  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/usehyptestforreliability",
2198  "should the strong branching decision be based on a hypothesis test?",
2199  &branchruledata->usehyptestforreliability, TRUE, DEFAULT_USEHYPTESTFORRELIABILITY,
2200  NULL, NULL) );
2201 
2202  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/usedynamicconfidence",
2203  "should the confidence level be adjusted dynamically?",
2204  &branchruledata->usedynamicconfidence, TRUE, DEFAULT_USEDYNAMICCONFIDENCE,
2205  NULL, NULL) );
2206  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/skipbadinitcands",
2207  "should branching rule skip candidates that have a low probability to "
2208  "be better than the best strong-branching or pseudo-candidate?",
2209  &branchruledata->skipbadinitcands, TRUE, DEFAULT_SKIPBADINITCANDS,
2210  NULL, NULL) );
2211  SCIP_CALL( SCIPaddIntParam(scip,
2212  "branching/relpscost/confidencelevel",
2213  "the confidence level for statistical methods, between 0 (Min) and 4 (Max).",
2214  &branchruledata->confidencelevel, TRUE, DEFAULT_CONFIDENCELEVEL, 0, 4, NULL, NULL) );
2215 
2216  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/randinitorder",
2217  "should candidates be initialized in randomized order?",
2218  &branchruledata->randinitorder, TRUE, DEFAULT_RANDINITORDER,
2219  NULL, NULL) );
2220 
2221  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/usesmallweightsitlim",
2222  "should smaller weights be used for pseudo cost updates after hitting the LP iteration limit?",
2223  &branchruledata->usesmallweightsitlim, TRUE, DEFAULT_USESMALLWEIGHTSITLIM,
2224  NULL, NULL) );
2225 
2226  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/dynamicweights",
2227  "should the weights of the branching rule be adjusted dynamically during solving based on objective and infeasible leaf counters?",
2228  &branchruledata->dynamicweights, TRUE, DEFAULT_DYNAMICWEIGHTS,
2229  NULL, NULL) );
2230  SCIP_CALL( SCIPaddIntParam(scip, "branching/relpscost/degeneracyaware",
2231  "should degeneracy be taken into account to update weights and skip strong branching? (0: off, 1: after root, 2: always)",
2232  &branchruledata->degeneracyaware, TRUE, DEFAULT_DEGENERACYAWARE, 0, 2,
2233  NULL, NULL) );
2234  SCIP_CALL( SCIPaddIntParam(scip, "branching/relpscost/startrandseed", "start seed for random number generation",
2235  &branchruledata->startrandseed, TRUE, DEFAULT_STARTRANDSEED, 0, INT_MAX, NULL, NULL) );
2236 
2237  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/filtercandssym",
2238  "Use symmetry to filter branching candidates?",
2239  &branchruledata->filtercandssym, TRUE, DEFAULT_FILTERCANDSSYM, NULL, NULL) );
2240 
2241  SCIP_CALL( SCIPaddBoolParam(scip, "branching/relpscost/transsympscost",
2242  "Transfer pscost information to symmetric variables?",
2243  &branchruledata->transsympscost, TRUE, DEFAULT_TRANSSYMPSCOST, NULL, NULL) );
2244 
2245  /* initialise the Treemodel parameters */
2246  SCIP_CALL( SCIPtreemodelInit(scip, &branchruledata->treemodel) );
2247 
2248  return SCIP_OKAY;
2249 }
2250 
2251 /** execution reliability pseudo cost branching with the given branching candidates */
2253  SCIP* scip, /**< SCIP data structure */
2254  SCIP_VAR** branchcands, /**< branching candidates */
2255  SCIP_Real* branchcandssol, /**< solution value for the branching candidates */
2256  SCIP_Real* branchcandsfrac, /**< fractional part of the branching candidates */
2257  int nbranchcands, /**< number of branching candidates */
2258  SCIP_Bool executebranching, /**< perform a branching step after probing */
2259  SCIP_RESULT* result /**< pointer to the result of the execution */
2260  )
2261 {
2262  SCIP_BRANCHRULE* branchrule;
2263 
2264  assert(scip != NULL);
2265  assert(result != NULL);
2266 
2267  /* find branching rule */
2268  branchrule = SCIPfindBranchrule(scip, BRANCHRULE_NAME);
2269  assert(branchrule != NULL);
2270 
2271  /* execute branching rule */
2272  SCIP_CALL( execRelpscost(scip, branchrule, branchcands, branchcandssol, branchcandsfrac, NULL, nbranchcands, executebranching, result) );
2273 
2274  return SCIP_OKAY;
2275 }
enum SCIP_Result SCIP_RESULT
Definition: type_result.h:61
SCIP_RETCODE SCIPgetLPBranchCands(SCIP *scip, SCIP_VAR ***lpcands, SCIP_Real **lpcandssol, SCIP_Real **lpcandsfrac, int *nlpcands, int *npriolpcands, int *nfracimplvars)
Definition: scip_branch.c:395
static SCIP_RETCODE countNonlinearities(SCIP *scip, int *nlcount, int nlcountsize, int *nlcountmax)
void SCIPfreeRandom(SCIP *scip, SCIP_RANDNUMGEN **randnumgen)
#define SCIPfreeBlockMemoryArray(scip, ptr, num)
Definition: scip_mem.h:110
#define DEFAULT_FILTERCANDSSYM
enum SCIP_BoundType SCIP_BOUNDTYPE
Definition: type_lp.h:59
static SCIP_Real calcNlscore(SCIP *scip, int *nlcount, int nlcountmax, int probindex)
reliable pseudo costs branching rule
#define SCIPreallocBlockMemoryArray(scip, ptr, oldnum, newnum)
Definition: scip_mem.h:99
SCIP_BRANCHRULEDATA * SCIPbranchruleGetData(SCIP_BRANCHRULE *branchrule)
Definition: branch.c:1849
#define NULL
Definition: def.h:267
SCIP_Real SCIPfeastol(SCIP *scip)
#define SCIPallocBlockMemoryArray(scip, ptr, num)
Definition: scip_mem.h:93
#define DEFAULT_PSCOSTWEIGHT
SCIP_Bool SCIPisNLPConstructed(SCIP *scip)
Definition: scip_nlp.c:110
static SCIP_RETCODE SCIPupdateVarPseudocostSymmetric(SCIP *scip, SCIP_BRANCHRULEDATA *branchruledata, SCIP_VAR *branchvar, int *branchorbitidx, int branchvaridx, SCIP_Real solvaldelta, SCIP_Real objdelta, SCIP_Real weight)
SCIP_RETCODE SCIPtightenVarLb(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound, SCIP_Bool force, SCIP_Bool *infeasible, SCIP_Bool *tightened)
Definition: scip_var.c:5205
static void freeBdchgs(SCIP *scip, int **bdchginds, SCIP_BOUNDTYPE **bdchgtypes, SCIP_Real **bdchgbounds, int *nbdchgs)
#define BRANCHRULE_MAXDEPTH
public methods for SCIP parameter handling
SCIP_NODE * SCIPgetCurrentNode(SCIP *scip)
Definition: scip_tree.c:91
public methods for branch and bound tree
SCIP_RETCODE SCIPgetBinvarRepresentative(SCIP *scip, SCIP_VAR *var, SCIP_VAR **repvar, SCIP_Bool *negated)
Definition: scip_var.c:1599
#define DEFAULT_MAXBDCHGS
SCIP_Bool SCIPisSumGE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
static SCIP_Real calcScore(SCIP *scip, SCIP_BRANCHRULEDATA *branchruledata, SCIP_Real conflictscore, SCIP_Real avgconflictscore, SCIP_Real conflengthscore, SCIP_Real avgconflengthscore, SCIP_Real inferencescore, SCIP_Real avginferencescore, SCIP_Real cutoffscore, SCIP_Real avgcutoffscore, SCIP_Real gmieffscore, SCIP_Real lastgmieffscore, SCIP_Real pscostscore, SCIP_Real avgpscostscore, SCIP_Real nlscore, SCIP_Real frac, SCIP_Real degeneracyfactor)
public methods for memory management
SCIP_CONSHDLR * SCIPfindConshdlr(SCIP *scip, const char *name)
Definition: scip_cons.c:941
SCIP_Real SCIPgetCutoffbound(SCIP *scip)
SCIP_Real SCIPnodeGetLowerbound(SCIP_NODE *node)
Definition: tree.c:7510
SCIP_RETCODE SCIPgetRealParam(SCIP *scip, const char *name, SCIP_Real *value)
Definition: scip_param.c:307
SCIP_Real SCIPgetVarPseudocostVal(SCIP *scip, SCIP_VAR *var, SCIP_Real solvaldelta)
Definition: scip_var.c:8816
SCIP_RETCODE SCIPgetSymmetry(SCIP *scip, int *npermvars, SCIP_VAR ***permvars, SCIP_HASHMAP **permvarmap, int *nperms, int ***perms, int ***permstrans, SCIP_Real *log10groupsize, SCIP_Bool *binvaraffected, int **components, int **componentbegins, int **vartocomponent, int *ncomponents)
#define DEFAULT_LOWERRORTOL
static long bound
#define DEFAULT_GMIAVGEFFWEIGHT
SCIP_RETCODE SCIPgetLPDualDegeneracy(SCIP *scip, SCIP_Real *degeneracy, SCIP_Real *varconsratio)
Definition: scip_lp.c:2792
SCIP_Real SCIPvarGetLbLocal(SCIP_VAR *var)
Definition: var.c:18135
SCIP_RETCODE SCIPgetVarStrongbranchLast(SCIP *scip, SCIP_VAR *var, SCIP_Real *down, SCIP_Real *up, SCIP_Bool *downvalid, SCIP_Bool *upvalid, SCIP_Real *solval, SCIP_Real *lpobjval)
Definition: scip_var.c:4012
SCIP_Bool SCIPisGE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
struct SCIP_BranchruleData SCIP_BRANCHRULEDATA
Definition: type_branch.h:57
#define DEFAULT_DEGENERACYAWARE
SCIP_Bool SCIPpscostThresholdProbabilityTest(SCIP *scip, SCIP_VAR *var, SCIP_Real frac, SCIP_Real threshold, SCIP_BRANCHDIR dir, SCIP_CONFIDENCELEVEL clevel)
Definition: scip_var.c:9061
static SCIP_RETCODE binvarGetActiveProbindex(SCIP *scip, SCIP_VAR *var, int *probindex)
SCIP_Real SCIPgetAvgConflictScore(SCIP *scip)
SCIP_RETCODE SCIPsetBranchruleInitsol(SCIP *scip, SCIP_BRANCHRULE *branchrule, SCIP_DECL_BRANCHINITSOL((*branchinitsol)))
Definition: scip_branch.c:217
SCIP_Bool SCIPvarIsBinary(SCIP_VAR *var)
Definition: var.c:17600
#define BRANCHRULE_MAXBOUNDDIST
#define DEFAULT_STARTRANDSEED
SCIP_Real SCIPgetVarPseudocostCountCurrentRun(SCIP *scip, SCIP_VAR *var, SCIP_BRANCHDIR dir)
Definition: scip_var.c:8952
SCIP_Real SCIPgetVarPseudocostScoreCurrentRun(SCIP *scip, SCIP_VAR *var, SCIP_Real solval)
Definition: scip_var.c:9143
SCIP_RETCODE SCIPsetBranchruleFree(SCIP *scip, SCIP_BRANCHRULE *branchrule, SCIP_DECL_BRANCHFREE((*branchfree)))
Definition: scip_branch.c:169
SCIP_RETCODE SCIPgetVarsData(SCIP *scip, SCIP_VAR ***vars, int *nvars, int *nbinvars, int *nintvars, int *nimplvars, int *ncontvars)
Definition: scip_prob.c:1866
#define DEFAULT_SBITERQUOT
SCIP_Real SCIPgetVarLastGMIScore(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:9934
SCIP_CONS ** SCIPconshdlrGetConss(SCIP_CONSHDLR *conshdlr)
Definition: cons.c:4595
#define DEFAULT_GMILASTEFFWEIGHT
#define FALSE
Definition: def.h:94
int SCIPgetSubscipDepth(SCIP *scip)
Definition: scip_copy.c:2605
SCIP_Real SCIPinfinity(SCIP *scip)
#define TRUE
Definition: def.h:93
#define SCIPdebug(x)
Definition: pub_message.h:93
SCIP_RETCODE SCIPsetBranchruleCopy(SCIP *scip, SCIP_BRANCHRULE *branchrule, SCIP_DECL_BRANCHCOPY((*branchcopy)))
Definition: scip_branch.c:153
enum SCIP_Retcode SCIP_RETCODE
Definition: type_retcode.h:63
Branching rules based on the Single-Variable-Branching (SVB) model.
SCIP_Longint SCIPgetVarStrongbranchNode(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:4162
#define DEFAULT_INFERENCEWEIGHT
SCIP_BRANCHRULE * SCIPfindBranchrule(SCIP *scip, const char *name)
Definition: scip_branch.c:297
int SCIPvarGetProbindex(SCIP_VAR *var)
Definition: var.c:17769
#define SCIP_UNUSED(x)
Definition: def.h:434
SCIP_Longint SCIPgetNRootStrongbranchLPIterations(SCIP *scip)
public methods for problem variables
SCIP_Bool SCIPisVarPscostRelerrorReliable(SCIP *scip, SCIP_VAR *var, SCIP_Real threshold, SCIP_CONFIDENCELEVEL clevel)
Definition: scip_var.c:9080
SCIP_RETCODE SCIPtightenVarUb(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound, SCIP_Bool force, SCIP_Bool *infeasible, SCIP_Bool *tightened)
Definition: scip_var.c:5322
#define SCIPfreeBlockMemory(scip, ptr)
Definition: scip_mem.h:108
public methods for branching rules
SCIP_RETCODE SCIPgetNLPVarsNonlinearity(SCIP *scip, int *nlcount)
Definition: scip_nlp.c:223
SCIP_RETCODE SCIPincludeBranchruleBasic(SCIP *scip, SCIP_BRANCHRULE **branchruleptr, const char *name, const char *desc, int priority, int maxdepth, SCIP_Real maxbounddist, SCIP_BRANCHRULEDATA *branchruledata)
Definition: scip_branch.c:116
Constraint handler for AND constraints, .
#define SCIPduplicateBufferArray(scip, ptr, source, num)
Definition: scip_mem.h:132
SCIP_Real SCIPfeasFrac(SCIP *scip, SCIP_Real val)
static SCIP_DECL_BRANCHINITSOL(branchInitsolRelpscost)
static SCIP_DECL_BRANCHFREE(branchFreeRelpscost)
#define SCIPfreeBufferArray(scip, ptr)
Definition: scip_mem.h:136
#define DEFAULT_HIGHERRORTOL
#define SCIPallocBlockMemory(scip, ptr)
Definition: scip_mem.h:89
public methods for SCIP variables
SCIP_VAR * SCIPvarGetNegationVar(SCIP_VAR *var)
Definition: var.c:17905
#define SCIPdebugMsg
Definition: scip_message.h:78
SCIP_RETCODE SCIPaddIntParam(SCIP *scip, const char *name, const char *desc, int *valueptr, SCIP_Bool isadvanced, int defaultvalue, int minvalue, int maxvalue, SCIP_DECL_PARAMCHGD((*paramchgd)), SCIP_PARAMDATA *paramdata)
Definition: scip_param.c:83
static SCIP_RETCODE initOrbits(SCIP *scip, SCIP_BRANCHRULEDATA *branchruledata)
SCIP_Real SCIPfeasCeil(SCIP *scip, SCIP_Real val)
static SCIP_DECL_BRANCHEXITSOL(branchExitsolRelpscost)
#define DEFAULT_USESMALLWEIGHTSITLIM
public methods for numerical tolerances
SCIP_RETCODE SCIPtreemodelSelectCandidate(SCIP *scip, SCIP_TREEMODEL *treemodel, SCIP_VAR **branchcands, SCIP_Real *mingains, SCIP_Real *maxgains, SCIP_Real *tiebreakerscore, int nbranchcands, int *bestcand)
Definition: treemodel.c:920
SCIP_Real SCIPgetVarPseudocostScore(SCIP *scip, SCIP_VAR *var, SCIP_Real solval)
Definition: scip_var.c:9105
SCIP_Real SCIPfeasFloor(SCIP *scip, SCIP_Real val)
SCIP_RETCODE SCIPsetBranchruleExecLp(SCIP *scip, SCIP_BRANCHRULE *branchrule, SCIP_DECL_BRANCHEXECLP((*branchexeclp)))
Definition: scip_branch.c:249
public methods for querying solving statistics
#define DEFAULT_INITITER
SCIP_Longint SCIPgetNDualResolveLPIterations(SCIP *scip)
SCIP_RETCODE SCIPupdateVarPseudocost(SCIP *scip, SCIP_VAR *var, SCIP_Real solvaldelta, SCIP_Real objdelta, SCIP_Real weight)
Definition: scip_var.c:8782
SCIP_Real SCIPgetBranchScore(SCIP *scip, SCIP_VAR *var, SCIP_Real downgain, SCIP_Real upgain)
Definition: scip_branch.c:849
public methods for the branch-and-bound tree
SCIP_Longint SCIPnodeGetNumber(SCIP_NODE *node)
Definition: tree.c:7490
#define BRANCHRULE_NAME
SCIP_Bool SCIPisLPSolBasic(SCIP *scip)
Definition: scip_lp.c:667
public methods for managing constraints
#define BRANCHRULE_PRIORITY
int SCIPgetNNLPVars(SCIP *scip)
Definition: scip_nlp.c:201
enum SCIP_Confidencelevel SCIP_CONFIDENCELEVEL
Definition: type_misc.h:53
SCIP_VAR ** SCIPgetVarsAnd(SCIP *scip, SCIP_CONS *cons)
Definition: cons_and.c:5230
static SCIP_RETCODE applyBdchgs(SCIP *scip, SCIP_VAR **vars, int *bdchginds, SCIP_BOUNDTYPE *bdchgtypes, SCIP_Real *bdchgbounds, int nbdchgs, SCIP_RESULT *result)
SCIP_VAR * SCIPgetResultantAnd(SCIP *scip, SCIP_CONS *cons)
Definition: cons_and.c:5255
SCIP_Bool SCIPisLT(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
#define SCIPfreeBufferArrayNull(scip, ptr)
Definition: scip_mem.h:137
const char * SCIPvarGetName(SCIP_VAR *var)
Definition: var.c:17420
#define DEFAULT_USEHYPTESTFORRELIABILITY
SCIP_Real SCIPgetVarConflictlengthScore(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:9305
SCIP_Real SCIPvarGetLPSol(SCIP_VAR *var)
Definition: var.c:18453
public methods for primal CIP solutions
#define SCIP_CALL(x)
Definition: def.h:380
SCIP_Bool SCIPsignificantVarPscostDifference(SCIP *scip, SCIP_VAR *varx, SCIP_Real fracx, SCIP_VAR *vary, SCIP_Real fracy, SCIP_BRANCHDIR dir, SCIP_CONFIDENCELEVEL clevel, SCIP_Bool onesided)
Definition: scip_var.c:9031
SCIP_RETCODE SCIPbranchVarVal(SCIP *scip, SCIP_VAR *var, SCIP_Real val, SCIP_NODE **downchild, SCIP_NODE **eqchild, SCIP_NODE **upchild)
Definition: scip_branch.c:1126
static SCIP_RETCODE filterSymmetricVariables(SCIP *scip, SCIP_BRANCHRULEDATA *branchruledata, SCIP_VAR **origbranchcands, SCIP_Real *origbranchcandssol, SCIP_Real *origbranchcandsfrac, int norigbranchcands, SCIP_VAR **branchcands, SCIP_Real *branchcandssol, SCIP_Real *branchcandsfrac, int *branchorbitidx, int *nbranchcands)
propagator for symmetry handling
void SCIPverbMessage(SCIP *scip, SCIP_VERBLEVEL msgverblevel, FILE *file, const char *formatstr,...)
Definition: scip_message.c:225
#define DEFAULT_SBITEROFS
#define DEFAULT_CUTOFFWEIGHT
SCIP_Bool SCIPtreemodelIsEnabled(SCIP *scip, SCIP_TREEMODEL *treemodel)
Definition: treemodel.c:908
SCIP_Bool SCIPhasCurrentNodeLP(SCIP *scip)
Definition: scip_lp.c:83
public methods for constraint handler plugins and constraints
SCIP_Longint SCIPgetNNodeLPIterations(SCIP *scip)
SCIP_Real SCIPgetVarPseudocostCurrentRun(SCIP *scip, SCIP_VAR *var, SCIP_BRANCHDIR dir)
Definition: scip_var.c:8898
#define DEFAULT_CONFLICTWEIGHT
SCIP_RETCODE SCIPcreateRandom(SCIP *scip, SCIP_RANDNUMGEN **randnumgen, unsigned int initialseed, SCIP_Bool useglobalseed)
static SCIP_RETCODE addBdchg(SCIP *scip, int **bdchginds, SCIP_BOUNDTYPE **bdchgtypes, SCIP_Real **bdchgbounds, int *nbdchgs, int ind, SCIP_BOUNDTYPE type, SCIP_Real bound)
#define SCIPallocBufferArray(scip, ptr, num)
Definition: scip_mem.h:124
SCIP_RETCODE SCIPstartStrongbranch(SCIP *scip, SCIP_Bool enablepropagation)
Definition: scip_var.c:2688
public data structures and miscellaneous methods
SCIP_Bool SCIPisSumGT(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
#define SCIP_Bool
Definition: def.h:91
SCIP_LPSOLSTAT SCIPgetLPSolstat(SCIP *scip)
Definition: scip_lp.c:168
static SCIP_DECL_BRANCHEXECLP(branchExeclpRelpscost)
static SCIP_DECL_BRANCHCOPY(branchCopyRelpscost)
#define DEFAULT_CONFLENGTHWEIGHT
#define DEFAULT_SKIPBADINITCANDS
int SCIPgetDepth(SCIP *scip)
Definition: scip_tree.c:670
SCIP_RETCODE SCIPupdateNodeLowerbound(SCIP *scip, SCIP_NODE *node, SCIP_Real newbound)
Definition: scip_prob.c:3757
SCIP_RETCODE SCIPtreemodelFree(SCIP *scip, SCIP_TREEMODEL **treemodel)
Definition: treemodel.c:892
void SCIPbranchruleSetData(SCIP_BRANCHRULE *branchrule, SCIP_BRANCHRULEDATA *branchruledata)
Definition: branch.c:1859
SCIP_RETCODE SCIPgetVarStrongbranchWithPropagation(SCIP *scip, SCIP_VAR *var, SCIP_Real solval, SCIP_Real lpobjval, int itlim, int maxproprounds, SCIP_Real *down, SCIP_Real *up, SCIP_Bool *downvalid, SCIP_Bool *upvalid, SCIP_Longint *ndomredsdown, SCIP_Longint *ndomredsup, SCIP_Bool *downinf, SCIP_Bool *upinf, SCIP_Bool *downconflict, SCIP_Bool *upconflict, SCIP_Bool *lperror, SCIP_Real *newlbs, SCIP_Real *newubs)
Definition: scip_var.c:3354
#define DEFAULT_MINRELIABLE
#define DEFAULT_MAXPROPROUNDS
#define MIN(x, y)
Definition: def.h:243
SCIP_RETCODE SCIPsetBranchruleExitsol(SCIP *scip, SCIP_BRANCHRULE *branchrule, SCIP_DECL_BRANCHEXITSOL((*branchexitsol)))
Definition: scip_branch.c:233
#define DEFAULT_INITCAND
SCIP_Longint SCIPgetNDualResolveLPs(SCIP *scip)
#define DEFAULT_NLSCOREWEIGHT
#define DEFAULT_DYNAMICWEIGHTS
#define DEFAULT_MAXLOOKAHEAD
SCIP_Longint SCIPgetNObjlimLeaves(SCIP *scip)
SCIP_RETCODE SCIPgetVarStrongbranchFrac(SCIP *scip, SCIP_VAR *var, int itlim, SCIP_Bool idempotent, SCIP_Real *down, SCIP_Real *up, SCIP_Bool *downvalid, SCIP_Bool *upvalid, SCIP_Bool *downinf, SCIP_Bool *upinf, SCIP_Bool *downconflict, SCIP_Bool *upconflict, SCIP_Bool *lperror)
Definition: scip_var.c:2921
int SCIPconshdlrGetNActiveConss(SCIP_CONSHDLR *conshdlr)
Definition: cons.c:4672
SCIP_Real SCIPrandomGetReal(SCIP_RANDNUMGEN *randnumgen, SCIP_Real minrandval, SCIP_Real maxrandval)
Definition: misc.c:10130
SCIP_Bool SCIPinProbing(SCIP *scip)
Definition: scip_probing.c:97
public methods for the LP relaxation, rows and columns
int SCIPgetNVars(SCIP *scip)
Definition: scip_prob.c:1992
#define SCIP_REAL_MAX
Definition: def.h:174
#define DEFAULT_USERELERRORFORRELIABILITY
public methods for nonlinear relaxation
static SCIP_RETCODE execRelpscost(SCIP *scip, SCIP_BRANCHRULE *branchrule, SCIP_VAR **branchcands, SCIP_Real *branchcandssol, SCIP_Real *branchcandsfrac, int *branchorbitidx, int nbranchcands, SCIP_Bool executebranch, SCIP_RESULT *result)
#define SCIP_REAL_MIN
Definition: def.h:175
SCIP_Bool SCIPinDive(SCIP *scip)
Definition: scip_lp.c:2775
#define SCIP_LONGINT_FORMAT
Definition: def.h:165
SCIP_Real SCIPgetAvgCutoffScore(SCIP *scip)
public methods for branching rule plugins and branching
SCIP_LPSOLSTAT SCIPgetLastStrongbranchLPSolStat(SCIP *scip, SCIP_BRANCHDIR branchdir)
Definition: scip_var.c:3990
SCIP_Longint SCIPgetNNodeInitLPIterations(SCIP *scip)
general public methods
SCIP_Longint SCIPgetNNodeInitLPs(SCIP *scip)
#define MAX(x, y)
Definition: def.h:239
SCIP_Real SCIPgetAvgConflictlengthScore(SCIP *scip)
SCIP_Real SCIPgetLPObjval(SCIP *scip)
Definition: scip_lp.c:247
SCIP_SOL * SCIPgetBestSol(SCIP *scip)
Definition: scip_sol.c:2169
SCIP_Bool SCIPisGT(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
#define DEFAULT_PROBINGBOUNDS
public methods for solutions
SCIP_Longint SCIPgetNStrongbranchLPIterations(SCIP *scip)
public methods for random numbers
methods for handling symmetries
SCIP_RETCODE SCIPincludeBranchruleRelpscost(SCIP *scip)
SCIP_RETCODE SCIPendStrongbranch(SCIP *scip)
Definition: scip_var.c:2746
#define DEFAULT_USESBLOCALINFO
public methods for message output
SCIP_RETCODE SCIPexecRelpscostBranching(SCIP *scip, SCIP_VAR **branchcands, SCIP_Real *branchcandssol, SCIP_Real *branchcandsfrac, int nbranchcands, SCIP_Bool executebranching, SCIP_RESULT *result)
#define DEFAULT_STORESEMIINITCOSTS
SCIP_RETCODE SCIPcomputeOrbitsComponentsSym(SCIP *scip, int npermvars, int **permstrans, int nperms, int *components, int *componentbegins, int *vartocomponent, int ncomponents, int *orbits, int *orbitbegins, int *norbits, int *varorbitmap)
Definition: symmetry.c:420
SCIP_VAR ** SCIPgetVars(SCIP *scip)
Definition: scip_prob.c:1947
SCIP_VARSTATUS SCIPvarGetStatus(SCIP_VAR *var)
Definition: var.c:17539
#define SCIP_Real
Definition: def.h:173
const char * SCIPbranchruleGetName(SCIP_BRANCHRULE *branchrule)
Definition: branch.c:1971
SCIP_Bool SCIPisStopped(SCIP *scip)
Definition: scip_general.c:718
SCIP_Real SCIPgetVarAvgGMIScore(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:9879
int SCIPgetNVarsAnd(SCIP *scip, SCIP_CONS *cons)
Definition: cons_and.c:5206
public methods for message handling
#define SCIP_INVALID
Definition: def.h:193
#define SCIP_Longint
Definition: def.h:158
SCIP_Bool SCIPallColsInLP(SCIP *scip)
Definition: scip_lp.c:649
#define DEFAULT_USEDYNAMICCONFIDENCE
static SCIP_RETCODE branchruledataEnsureNlcount(SCIP *scip, SCIP_BRANCHRULEDATA *branchruledata)
#define DEFAULT_TRANSSYMPSCOST
SCIP_Real SCIPgetNodeLowerbound(SCIP *scip, SCIP_NODE *node)
Definition: scip_prob.c:3622
SCIP_Bool SCIPisZero(SCIP *scip, SCIP_Real val)
#define BRANCHRULE_DESC
SCIP_Real SCIPgetAvgPseudocostScore(SCIP *scip)
SCIP_Real SCIPvarGetUbLocal(SCIP_VAR *var)
Definition: var.c:18145
#define SCIPfreeBlockMemoryArrayNull(scip, ptr, num)
Definition: scip_mem.h:111
SCIP_Bool SCIPisFeasIntegral(SCIP *scip, SCIP_Real val)
#define BMSclearMemoryArray(ptr, num)
Definition: memory.h:130
int SCIPhashmapGetImageInt(SCIP_HASHMAP *hashmap, void *origin)
Definition: misc.c:3281
SCIP_RETCODE SCIPtreemodelInit(SCIP *scip, SCIP_TREEMODEL **treemodel)
Definition: treemodel.c:834
#define DEFAULT_MAXRELIABLE
SCIP_Longint SCIPgetNNodes(SCIP *scip)
public methods for global and local (sub)problems
SCIP_Longint SCIPgetNInfeasibleLeaves(SCIP *scip)
#define DEFAULT_CONFIDENCELEVEL
#define DEFAULT_RANDINITORDER
SCIP_Real SCIPgetVarAvgInferenceScore(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:9475
SCIP_Bool SCIPisExactSolve(SCIP *scip)
Definition: scip_general.c:605
int SCIPsolGetIndex(SCIP_SOL *sol)
Definition: sol.c:2835
SCIP_RETCODE SCIPaddRealParam(SCIP *scip, const char *name, const char *desc, SCIP_Real *valueptr, SCIP_Bool isadvanced, SCIP_Real defaultvalue, SCIP_Real minvalue, SCIP_Real maxvalue, SCIP_DECL_PARAMCHGD((*paramchgd)), SCIP_PARAMDATA *paramdata)
Definition: scip_param.c:139
SCIP_Real SCIPgetAvgInferenceScore(SCIP *scip)
SCIP_RETCODE SCIPaddBoolParam(SCIP *scip, const char *name, const char *desc, SCIP_Bool *valueptr, SCIP_Bool isadvanced, SCIP_Bool defaultvalue, SCIP_DECL_PARAMCHGD((*paramchgd)), SCIP_PARAMDATA *paramdata)
Definition: scip_param.c:57
SCIP_Real SCIPgetVarConflictScore(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:9243
SCIP_Bool SCIPvarIsActive(SCIP_VAR *var)
Definition: var.c:17749
SCIP_Bool SCIPvarIsNegated(SCIP_VAR *var)
Definition: var.c:17575
#define SCIPreallocBufferArray(scip, ptr, num)
Definition: scip_mem.h:128
memory allocation routines
SCIP_Real SCIPgetVarAvgCutoffScore(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:9729