xref: /petsc/src/tao/bound/impls/bncg/bncg.c (revision 7529f6b4a072b190be86ef0661bc765cc410dd40) 
1 #include <petsctaolinesearch.h>
2 #include <../src/tao/bound/impls/bncg/bncg.h>
3 
4 #define CG_FletcherReeves       0
5 #define CG_PolakRibiere         1
6 #define CG_PolakRibierePlus     2
7 #define CG_HestenesStiefel      3
8 #define CG_DaiYuan              4
9 #define CG_Types                5
10 
11 static const char *CG_Table[64] = {"fr", "pr", "prp", "hs", "dy"};
12 
13 #define CG_AS_NONE       0
14 #define CG_AS_BERTSEKAS  1
15 #define CG_AS_SIZE       2
16 
17 static const char *CG_AS_TYPE[64] = {"none", "bertsekas"};
18 
19 PetscErrorCode TaoBNCGSetRecycleFlag(Tao tao, PetscBool recycle)
20 {
21   TAO_BNCG                     *cg = (TAO_BNCG*)tao->data;
22 
23   PetscFunctionBegin;
24   cg->recycle = recycle;
25   PetscFunctionReturn(0);
26 }
27 
28 PetscErrorCode TaoBNCGEstimateActiveSet(Tao tao, PetscInt asType)
29 {
30   PetscErrorCode               ierr;
31   TAO_BNCG                     *cg = (TAO_BNCG *)tao->data;
32 
33   PetscFunctionBegin;
34   ierr = ISDestroy(&cg->inactive_old);CHKERRQ(ierr);
35   if (cg->inactive_idx) {
36     ierr = ISDuplicate(cg->inactive_idx, &cg->inactive_old);CHKERRQ(ierr);
37     ierr = ISCopy(cg->inactive_idx, cg->inactive_old);CHKERRQ(ierr);
38   }
39   switch (asType) {
40   case CG_AS_NONE:
41     ierr = ISDestroy(&cg->inactive_idx);CHKERRQ(ierr);
42     ierr = VecWhichInactive(tao->XL, tao->solution, cg->unprojected_gradient, tao->XU, PETSC_TRUE, &cg->inactive_idx);CHKERRQ(ierr);
43     ierr = ISDestroy(&cg->active_idx);CHKERRQ(ierr);
44     ierr = ISComplementVec(cg->inactive_idx, tao->solution, &cg->active_idx);CHKERRQ(ierr);
45     break;
46 
47   case CG_AS_BERTSEKAS:
48     /* Use gradient descent to estimate the active set */
49     ierr = VecCopy(cg->unprojected_gradient, cg->W);CHKERRQ(ierr);
50     ierr = VecScale(cg->W, -1.0);CHKERRQ(ierr);
51     ierr = TaoEstimateActiveBounds(tao->solution, tao->XL, tao->XU, cg->unprojected_gradient, cg->W, cg->work, cg->as_step, &cg->as_tol, &cg->active_lower, &cg->active_upper, &cg->active_fixed, &cg->active_idx, &cg->inactive_idx);CHKERRQ(ierr);
52     break;
53 
54   default:
55     break;
56   }
57   PetscFunctionReturn(0);
58 }
59 
60 PetscErrorCode TaoBNCGBoundStep(Tao tao, Vec step)
61 {
62   PetscErrorCode               ierr;
63   TAO_BNCG                     *cg = (TAO_BNCG *)tao->data;
64 
65   PetscFunctionBegin;
66   switch (cg->as_type) {
67   case CG_AS_NONE:
68     if (cg->active_idx) {
69       ierr = VecISSet(step, cg->active_idx, 0.0);CHKERRQ(ierr);
70     }
71     break;
72 
73   case CG_AS_BERTSEKAS:
74     ierr = TaoBoundStep(tao->solution, tao->XL, tao->XU, cg->active_lower, cg->active_upper, cg->active_fixed, 1.0, step);CHKERRQ(ierr);
75     break;
76 
77   default:
78     break;
79   }
80   PetscFunctionReturn(0);
81 }
82 
83 static PetscErrorCode TaoSolve_BNCG(Tao tao)
84 {
85   TAO_BNCG                     *cg = (TAO_BNCG*)tao->data;
86   PetscErrorCode               ierr;
87   TaoLineSearchConvergedReason ls_status = TAOLINESEARCH_CONTINUE_ITERATING;
88   PetscReal                    step=1.0,gnorm,gnorm2,gd,ginner,beta,dnorm,resnorm;
89   PetscReal                    gd_old,gnorm2_old,f_old;
90   PetscBool                    cg_restart;
91   PetscInt                     nDiff;
92 
93   PetscFunctionBegin;
94   /*   Project the current point onto the feasible set */
95   ierr = TaoComputeVariableBounds(tao);CHKERRQ(ierr);
96   ierr = TaoLineSearchSetVariableBounds(tao->linesearch,tao->XL,tao->XU);CHKERRQ(ierr);
97 
98   /* Project the initial point onto the feasible region */
99   ierr = TaoBoundSolution(tao->XL,tao->XU,tao->solution, &nDiff);CHKERRQ(ierr);
100 
101   if (!cg->recycle) {
102     /*  Solver is not being recycled so just compute the objective function and criteria */
103     ierr = TaoComputeObjectiveAndGradient(tao, tao->solution, &cg->f, cg->unprojected_gradient);CHKERRQ(ierr);
104   } else {
105     /* We are recycling, so we have to compute ||g_old||^2 for use in the CG step calculation */
106     ierr = VecDot(cg->G_old, cg->G_old, &gnorm2_old);CHKERRQ(ierr);
107   }
108   ierr = VecNorm(cg->unprojected_gradient,NORM_2,&gnorm);CHKERRQ(ierr);
109   if (PetscIsInfOrNanReal(cg->f) || PetscIsInfOrNanReal(gnorm)) SETERRQ(PETSC_COMM_SELF,1, "User provided compute function generated Inf or NaN");
110 
111   /* Estimate the active set and compute the projected gradient */
112   ierr = VecCopy(cg->unprojected_gradient, cg->W);CHKERRQ(ierr);
113   ierr = VecScale(cg->W, -1.0);CHKERRQ(ierr);
114   ierr = TaoBNCGEstimateActiveSet(tao, cg->as_type);CHKERRQ(ierr);
115 
116   /* Project the gradient and calculate the norm */
117   ierr = VecCopy(cg->unprojected_gradient, tao->gradient);CHKERRQ(ierr);
118   if (cg->active_idx) {
119     ierr = VecISSet(tao->gradient, cg->active_idx, 0.0);CHKERRQ(ierr);
120   }
121   ierr = VecNorm(tao->gradient,NORM_2,&gnorm);CHKERRQ(ierr);
122   gnorm2 = gnorm*gnorm;
123 
124   /* Convergence check */
125   tao->niter = 0;
126   tao->reason = TAO_CONTINUE_ITERATING;
127   ierr = VecFischer(tao->solution, cg->unprojected_gradient, tao->XL, tao->XU, cg->W);CHKERRQ(ierr);
128   ierr = VecNorm(cg->W, NORM_2, &resnorm);CHKERRQ(ierr);
129   ierr = TaoLogConvergenceHistory(tao, cg->f, resnorm, 0.0, tao->ksp_its);CHKERRQ(ierr);
130   ierr = TaoMonitor(tao, tao->niter, cg->f, resnorm, 0.0, step);CHKERRQ(ierr);
131   ierr = (*tao->ops->convergencetest)(tao,tao->cnvP);CHKERRQ(ierr);
132   if (tao->reason != TAO_CONTINUE_ITERATING) PetscFunctionReturn(0);
133 
134   /* Start optimization iterations */
135   cg->ls_fails = cg->broken_ortho = cg->descent_error = 0;
136   cg->resets = -1;
137   while (tao->reason == TAO_CONTINUE_ITERATING) {
138     /* Check restart conditions for using steepest descent */
139     ++tao->niter;
140     cg_restart = PETSC_FALSE;
141     ierr = VecDot(tao->gradient, cg->G_old, &ginner);CHKERRQ(ierr);
142     ierr = VecNorm(tao->stepdirection, NORM_2, &dnorm);CHKERRQ(ierr);
143     if (tao->niter == 1 && !cg->recycle && dnorm != 0.0) {
144       /* 1) First iteration, with recycle disabled, and a non-zero previous step */
145       cg_restart = PETSC_TRUE;
146     } else if (PetscAbsScalar(ginner) >= cg->eta * gnorm2) {
147       /* 2) Gradients are far from orthogonal */
148       cg_restart = PETSC_TRUE;
149       ++cg->broken_ortho;
150     }
151 
152     /* Compute CG step */
153     if (cg_restart) {
154       beta = 0.0;
155       ++cg->resets;
156     } else {
157       switch (cg->cg_type) {
158       case CG_FletcherReeves:
159         beta = gnorm2 / gnorm2_old;
160         break;
161 
162       case CG_PolakRibiere:
163         beta = (gnorm2 - ginner) / gnorm2_old;
164         break;
165 
166       case CG_PolakRibierePlus:
167         beta = PetscMax((gnorm2-ginner)/gnorm2_old, 0.0);
168         break;
169 
170       case CG_HestenesStiefel:
171         ierr = VecDot(tao->gradient, tao->stepdirection, &gd);CHKERRQ(ierr);
172         ierr = VecDot(cg->G_old, tao->stepdirection, &gd_old);CHKERRQ(ierr);
173         beta = (gnorm2 - ginner) / (gd - gd_old);
174         break;
175 
176       case CG_DaiYuan:
177         ierr = VecDot(tao->gradient, tao->stepdirection, &gd);CHKERRQ(ierr);
178         ierr = VecDot(cg->G_old, tao->stepdirection, &gd_old);CHKERRQ(ierr);
179         beta = gnorm2 / (gd - gd_old);
180         break;
181 
182       default:
183         beta = 0.0;
184         break;
185       }
186     }
187 
188     /*  Compute the direction d=-g + beta*d */
189     ierr = VecAXPBY(tao->stepdirection, -1.0, beta, tao->gradient);CHKERRQ(ierr);
190     ierr = TaoBNCGBoundStep(tao, tao->stepdirection);CHKERRQ(ierr);
191     if (cg->inactive_old) {
192       /* Compute which new indexes that were active before became inactive this iteration */
193       ierr = ISDestroy(&cg->new_inactives);CHKERRQ(ierr);
194       ierr = ISDifference(cg->inactive_old, cg->inactive_idx, &cg->new_inactives);
195       /* Selectively reset the CG step those freshly inactive variables to be the gradient descent direction */
196       if (cg->new_inactives) {
197         ierr = VecGetSubVector(tao->stepdirection, cg->new_inactives, &cg->inactive_step);CHKERRQ(ierr);
198         ierr = VecGetSubVector(tao->gradient, cg->new_inactives, &cg->inactive_grad);CHKERRQ(ierr);
199         ierr = VecCopy(cg->inactive_grad, cg->inactive_step);CHKERRQ(ierr);
200         ierr = VecScale(cg->inactive_step, -1.0);CHKERRQ(ierr);
201         ierr = VecRestoreSubVector(tao->stepdirection, cg->new_inactives, &cg->inactive_step);CHKERRQ(ierr);
202         ierr = VecRestoreSubVector(tao->gradient, cg->new_inactives, &cg->inactive_grad);CHKERRQ(ierr);
203       }
204     }
205 
206     /* Verify that this is a descent direction */
207     ierr = VecDot(tao->gradient, tao->stepdirection, &gd);CHKERRQ(ierr);
208     ierr = VecNorm(tao->stepdirection, NORM_2, &dnorm);
209     if (gd > -cg->rho*PetscPowReal(dnorm, cg->pow)) {
210       /* Not a descent direction, so we reset back to projected gradient descent */
211       ierr = VecAXPBY(tao->stepdirection, -1.0, 0.0, tao->gradient);CHKERRQ(ierr);
212       ++cg->resets;
213       ++cg->descent_error;
214     }
215 
216     /* Store solution and gradient info before it changes */
217     ierr = VecCopy(tao->solution, cg->X_old);CHKERRQ(ierr);
218     ierr = VecCopy(tao->gradient, cg->G_old);CHKERRQ(ierr);
219     ierr = VecCopy(cg->unprojected_gradient, cg->unprojected_gradient_old);CHKERRQ(ierr);
220     gnorm2_old = gnorm2;
221     f_old = cg->f;
222 
223     /* Perform bounded line search */
224     step = 1.0;
225     ierr = TaoLineSearchApply(tao->linesearch, tao->solution, &cg->f, cg->unprojected_gradient, tao->stepdirection, &step, &ls_status);CHKERRQ(ierr);
226     ierr = TaoAddLineSearchCounts(tao);CHKERRQ(ierr);
227 
228     /*  Check linesearch failure */
229     if (ls_status != TAOLINESEARCH_SUCCESS && ls_status != TAOLINESEARCH_SUCCESS_USER) {
230       ++cg->ls_fails;
231       /* Restore previous point */
232       gnorm2 = gnorm2_old;
233       cg->f = f_old;
234       ierr = VecCopy(cg->X_old, tao->solution);CHKERRQ(ierr);
235       ierr = VecCopy(cg->G_old, tao->gradient);CHKERRQ(ierr);
236       ierr = VecCopy(cg->unprojected_gradient_old, cg->unprojected_gradient);CHKERRQ(ierr);
237 
238       /* Fall back on the gradient descent step */
239       ierr = VecCopy(tao->gradient, tao->stepdirection);CHKERRQ(ierr);
240       ierr = VecScale(tao->stepdirection, -1.0);CHKERRQ(ierr);
241       ierr = TaoBNCGBoundStep(tao, tao->stepdirection);CHKERRQ(ierr);
242 
243       step = 1.0;
244       ierr = TaoLineSearchApply(tao->linesearch, tao->solution, &cg->f, cg->unprojected_gradient, tao->stepdirection, &step, &ls_status);CHKERRQ(ierr);
245       ierr = TaoAddLineSearchCounts(tao);CHKERRQ(ierr);
246 
247       if (ls_status != TAOLINESEARCH_SUCCESS && ls_status != TAOLINESEARCH_SUCCESS_USER){
248         ++cg->ls_fails;
249         /* Restore previous point */
250         gnorm2 = gnorm2_old;
251         cg->f = f_old;
252         ierr = VecCopy(cg->X_old, tao->solution);CHKERRQ(ierr);
253         ierr = VecCopy(cg->G_old, tao->gradient);CHKERRQ(ierr);
254         ierr = VecCopy(cg->unprojected_gradient_old, cg->unprojected_gradient);CHKERRQ(ierr);
255 
256         /* Nothing left to do but fail out of the optimization */
257         step = 0.0;
258         tao->reason = TAO_DIVERGED_LS_FAILURE;
259       }
260     }
261 
262     if (tao->reason != TAO_DIVERGED_LS_FAILURE) {
263       /* Estimate the active set at the new solution */
264       ierr = TaoBNCGEstimateActiveSet(tao, cg->as_type);CHKERRQ(ierr);
265 
266       /* Compute the projected gradient and its norm */
267       ierr = VecCopy(cg->unprojected_gradient, tao->gradient);CHKERRQ(ierr);
268       if (cg->active_idx) {
269         ierr = VecISSet(tao->gradient, cg->active_idx, 0.0);CHKERRQ(ierr);
270       }
271       ierr = VecNorm(tao->gradient,NORM_2,&gnorm);CHKERRQ(ierr);
272       gnorm2 = gnorm*gnorm;
273     }
274 
275     /* Convergence test */
276     ierr = VecFischer(tao->solution, cg->unprojected_gradient, tao->XL, tao->XU, cg->W);CHKERRQ(ierr);
277     ierr = VecNorm(cg->W, NORM_2, &resnorm);CHKERRQ(ierr);
278     ierr = TaoLogConvergenceHistory(tao, cg->f, resnorm, 0.0, tao->ksp_its);CHKERRQ(ierr);
279     ierr = TaoMonitor(tao, tao->niter, cg->f, resnorm, 0.0, step);CHKERRQ(ierr);
280     ierr = (*tao->ops->convergencetest)(tao,tao->cnvP);CHKERRQ(ierr);
281   }
282   PetscFunctionReturn(0);
283 }
284 
285 static PetscErrorCode TaoSetUp_BNCG(Tao tao)
286 {
287   TAO_BNCG         *cg = (TAO_BNCG*)tao->data;
288   PetscErrorCode ierr;
289 
290   PetscFunctionBegin;
291   if (!tao->gradient) {
292     ierr = VecDuplicate(tao->solution,&tao->gradient);CHKERRQ(ierr);
293   }
294   if (!tao->stepdirection) {
295     ierr = VecDuplicate(tao->solution,&tao->stepdirection);CHKERRQ(ierr);
296   }
297   if (!cg->W) {
298     ierr = VecDuplicate(tao->solution,&cg->W);CHKERRQ(ierr);
299   }
300   if (!cg->work) {
301     ierr = VecDuplicate(tao->solution,&cg->work);CHKERRQ(ierr);
302   }
303   if (!cg->X_old) {
304     ierr = VecDuplicate(tao->solution,&cg->X_old);CHKERRQ(ierr);
305   }
306   if (!cg->G_old) {
307     ierr = VecDuplicate(tao->gradient,&cg->G_old);CHKERRQ(ierr);
308   }
309   if (!cg->unprojected_gradient) {
310     ierr = VecDuplicate(tao->gradient,&cg->unprojected_gradient);CHKERRQ(ierr);
311   }
312   if (!cg->unprojected_gradient_old) {
313     ierr = VecDuplicate(tao->gradient,&cg->unprojected_gradient_old);CHKERRQ(ierr);
314   }
315   PetscFunctionReturn(0);
316 }
317 
318 static PetscErrorCode TaoDestroy_BNCG(Tao tao)
319 {
320   TAO_BNCG       *cg = (TAO_BNCG*) tao->data;
321   PetscErrorCode ierr;
322 
323   PetscFunctionBegin;
324   if (tao->setupcalled) {
325     ierr = VecDestroy(&cg->W);CHKERRQ(ierr);
326     ierr = VecDestroy(&cg->work);CHKERRQ(ierr);
327     ierr = VecDestroy(&cg->X_old);CHKERRQ(ierr);
328     ierr = VecDestroy(&cg->G_old);CHKERRQ(ierr);
329     ierr = VecDestroy(&cg->unprojected_gradient);CHKERRQ(ierr);
330     ierr = VecDestroy(&cg->unprojected_gradient_old);CHKERRQ(ierr);
331   }
332   ierr = PetscFree(tao->data);CHKERRQ(ierr);
333   PetscFunctionReturn(0);
334 }
335 
336 static PetscErrorCode TaoSetFromOptions_BNCG(PetscOptionItems *PetscOptionsObject,Tao tao)
337  {
338     TAO_BNCG       *cg = (TAO_BNCG*)tao->data;
339     PetscErrorCode ierr;
340 
341     PetscFunctionBegin;
342     ierr = TaoLineSearchSetFromOptions(tao->linesearch);CHKERRQ(ierr);
343     ierr = PetscOptionsHead(PetscOptionsObject,"Nonlinear Conjugate Gradient method for unconstrained optimization");CHKERRQ(ierr);
344     ierr = PetscOptionsReal("-tao_bncg_eta","restart tolerance", "", cg->eta,&cg->eta,NULL);CHKERRQ(ierr);
345     ierr = PetscOptionsReal("-tao_bncg_rho","descent direction tolerance", "", cg->rho,&cg->rho,NULL);CHKERRQ(ierr);
346     ierr = PetscOptionsReal("-tao_bncg_pow","descent direction exponent", "", cg->pow,&cg->pow,NULL);CHKERRQ(ierr);
347     ierr = PetscOptionsEList("-tao_bncg_type","cg formula", "", CG_Table, CG_Types, CG_Table[cg->cg_type], &cg->cg_type,NULL);CHKERRQ(ierr);
348     ierr = PetscOptionsEList("-tao_bncg_as_type","active set estimation method", "", CG_AS_TYPE, CG_AS_SIZE, CG_AS_TYPE[cg->cg_type], &cg->cg_type,NULL);CHKERRQ(ierr);
349     ierr = PetscOptionsReal("-tao_bncg_delta_min","minimum delta value", "", cg->delta_min,&cg->delta_min,NULL);CHKERRQ(ierr);
350     ierr = PetscOptionsReal("-tao_bncg_delta_max","maximum delta value", "", cg->delta_max,&cg->delta_max,NULL);CHKERRQ(ierr);
351     ierr = PetscOptionsBool("-tao_bncg_recycle","enable recycling the existing solution and gradient at the start of a new solve","",cg->recycle,&cg->recycle,NULL);CHKERRQ(ierr);
352     ierr = PetscOptionsReal("-tao_bncg_as_tol", "initial tolerance used when estimating actively bounded variables","",cg->as_tol,&cg->as_tol,NULL);CHKERRQ(ierr);
353     ierr = PetscOptionsReal("-tao_bncg_as_step", "step length used when estimating actively bounded variables","",cg->as_step,&cg->as_step,NULL);CHKERRQ(ierr);
354    ierr = PetscOptionsTail();CHKERRQ(ierr);
355    PetscFunctionReturn(0);
356 }
357 
358 static PetscErrorCode TaoView_BNCG(Tao tao, PetscViewer viewer)
359 {
360   PetscBool      isascii;
361   TAO_BNCG       *cg = (TAO_BNCG*)tao->data;
362   PetscErrorCode ierr;
363 
364   PetscFunctionBegin;
365   ierr = PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii);CHKERRQ(ierr);
366   if (isascii) {
367     ierr = PetscViewerASCIIPushTab(viewer);CHKERRQ(ierr);
368     ierr = PetscViewerASCIIPrintf(viewer, "CG Type: %s\n", CG_Table[cg->cg_type]);CHKERRQ(ierr);
369     ierr = PetscViewerASCIIPrintf(viewer, "Resets: %i\n", cg->resets);CHKERRQ(ierr);
370     ierr = PetscViewerASCIIPrintf(viewer, "  Broken ortho: %i\n", cg->broken_ortho);CHKERRQ(ierr);
371     ierr = PetscViewerASCIIPrintf(viewer, "  Not a descent dir.: %i\n", cg->descent_error);CHKERRQ(ierr);
372     ierr = PetscViewerASCIIPrintf(viewer, "Line search fails: %i\n", cg->ls_fails);CHKERRQ(ierr);
373     ierr = PetscViewerASCIIPopTab(viewer);CHKERRQ(ierr);
374   }
375   PetscFunctionReturn(0);
376 }
377 
378 /*MC
379      TAOBNCG -   Bound-constrained Nonlinear Conjugate Gradient method.
380 
381    Options Database Keys:
382 +      -tao_bncg_recycle - enable recycling the latest calculated gradient vector in subsequent TaoSolve() calls
383 .      -tao_bncg_eta <r> - restart tolerance
384 .      -tao_bncg_type <taocg_type> - cg formula
385 .      -tao_bncg_as_type <none,bertsekas> - active set estimation method
386 .      -tao_bncg_as_tol <r> - tolerance used in Bertsekas active-set estimation
387 .      -tao_bncg_as_step <r> - trial step length used in Bertsekas active-set estimation
388 .      -tao_bncg_delta_min <r> - minimum delta value
389 -      -tao_bncg_delta_max <r> - maximum delta value
390 
391   Notes:
392      CG formulas are:
393          "fr" - Fletcher-Reeves
394          "pr" - Polak-Ribiere
395          "prp" - Polak-Ribiere-Plus
396          "hs" - Hestenes-Steifel
397          "dy" - Dai-Yuan
398   Level: beginner
399 M*/
400 
401 
402 PETSC_EXTERN PetscErrorCode TaoCreate_BNCG(Tao tao)
403 {
404   TAO_BNCG       *cg;
405   const char     *morethuente_type = TAOLINESEARCHMT;
406   PetscErrorCode ierr;
407 
408   PetscFunctionBegin;
409   tao->ops->setup = TaoSetUp_BNCG;
410   tao->ops->solve = TaoSolve_BNCG;
411   tao->ops->view = TaoView_BNCG;
412   tao->ops->setfromoptions = TaoSetFromOptions_BNCG;
413   tao->ops->destroy = TaoDestroy_BNCG;
414 
415   /* Override default settings (unless already changed) */
416   if (!tao->max_it_changed) tao->max_it = 2000;
417   if (!tao->max_funcs_changed) tao->max_funcs = 4000;
418 
419   /*  Note: nondefault values should be used for nonlinear conjugate gradient  */
420   /*  method.  In particular, gtol should be less that 0.5; the value used in  */
421   /*  Nocedal and Wright is 0.10.  We use the default values for the  */
422   /*  linesearch because it seems to work better. */
423   ierr = TaoLineSearchCreate(((PetscObject)tao)->comm, &tao->linesearch);CHKERRQ(ierr);
424   ierr = PetscObjectIncrementTabLevel((PetscObject)tao->linesearch, (PetscObject)tao, 1);CHKERRQ(ierr);
425   ierr = TaoLineSearchSetType(tao->linesearch, morethuente_type);CHKERRQ(ierr);
426   ierr = TaoLineSearchUseTaoRoutines(tao->linesearch, tao);CHKERRQ(ierr);
427   ierr = TaoLineSearchSetOptionsPrefix(tao->linesearch,tao->hdr.prefix);CHKERRQ(ierr);
428 
429   ierr = PetscNewLog(tao,&cg);CHKERRQ(ierr);
430   tao->data = (void*)cg;
431   cg->rho = 1e-4;
432   cg->pow = 2.1;
433   cg->eta = 0.5;
434   cg->delta_min = 1e-7;
435   cg->delta_max = 100;
436   cg->as_step = 0.001;
437   cg->as_tol = 0.001;
438   cg->as_type = CG_AS_BERTSEKAS;
439   cg->cg_type = CG_DaiYuan;
440   cg->recycle = PETSC_FALSE;
441   PetscFunctionReturn(0);
442 }
443