xref: /petsc/src/tao/complementarity/impls/asls/asils.c (revision ebead697dbf761eb322f829370bbe90b3bd93fa3)
1 #include <../src/tao/complementarity/impls/ssls/ssls.h>
2 /*
3    Context for ASXLS
4      -- active-set      - reduced matrices formed
5                           - inherit properties of original system
6      -- semismooth (S)  - function not differentiable
7                         - merit function continuously differentiable
8                         - Fischer-Burmeister reformulation of complementarity
9                           - Billups composition for two finite bounds
10      -- infeasible (I)  - iterates not guaranteed to remain within bounds
11      -- feasible (F)    - iterates guaranteed to remain within bounds
12      -- linesearch (LS) - Armijo rule on direction
13 
14    Many other reformulations are possible and combinations of
15    feasible/infeasible and linesearch/trust region are possible.
16 
17    Basic theory
18      Fischer-Burmeister reformulation is semismooth with a continuously
19      differentiable merit function and strongly semismooth if the F has
20      lipschitz continuous derivatives.
21 
22      Every accumulation point generated by the algorithm is a stationary
23      point for the merit function.  Stationary points of the merit function
24      are solutions of the complementarity problem if
25        a.  the stationary point has a BD-regular subdifferential, or
26        b.  the Schur complement F'/F'_ff is a P_0-matrix where ff is the
27            index set corresponding to the free variables.
28 
29      If one of the accumulation points has a BD-regular subdifferential then
30        a.  the entire sequence converges to this accumulation point at
31            a local q-superlinear rate
32        b.  if in addition the reformulation is strongly semismooth near
33            this accumulation point, then the algorithm converges at a
34            local q-quadratic rate.
35 
36    The theory for the feasible version follows from the feasible descent
37    algorithm framework.
38 
39    References:
40 +  * - Billups, "Algorithms for Complementarity Problems and Generalized
41        Equations," Ph.D thesis, University of Wisconsin  Madison, 1995.
42 .  * - De Luca, Facchinei, Kanzow, "A Semismooth Equation Approach to the
43        Solution of Nonlinear Complementarity Problems," Mathematical
44        Programming, 75, 1996.
45 .  * - Ferris, Kanzow, Munson, "Feasible Descent Algorithms for Mixed
46        Complementarity Problems," Mathematical Programming, 86,
47        1999.
48 .  * - Fischer, "A Special Newton type Optimization Method," Optimization,
49        24, 1992
50 -  * - Munson, Facchinei, Ferris, Fischer, Kanzow, "The Semismooth Algorithm
51        for Large Scale Complementarity Problems," Technical Report,
52        University of Wisconsin  Madison, 1999.
53 */
54 
55 static PetscErrorCode TaoSetUp_ASILS(Tao tao)
56 {
57   TAO_SSLS       *asls = (TAO_SSLS *)tao->data;
58 
59   PetscFunctionBegin;
60   PetscCall(VecDuplicate(tao->solution,&tao->gradient));
61   PetscCall(VecDuplicate(tao->solution,&tao->stepdirection));
62   PetscCall(VecDuplicate(tao->solution,&asls->ff));
63   PetscCall(VecDuplicate(tao->solution,&asls->dpsi));
64   PetscCall(VecDuplicate(tao->solution,&asls->da));
65   PetscCall(VecDuplicate(tao->solution,&asls->db));
66   PetscCall(VecDuplicate(tao->solution,&asls->t1));
67   PetscCall(VecDuplicate(tao->solution,&asls->t2));
68   asls->fixed = NULL;
69   asls->free = NULL;
70   asls->J_sub = NULL;
71   asls->Jpre_sub = NULL;
72   asls->w = NULL;
73   asls->r1 = NULL;
74   asls->r2 = NULL;
75   asls->r3 = NULL;
76   asls->dxfree = NULL;
77   PetscFunctionReturn(0);
78 }
79 
80 static PetscErrorCode Tao_ASLS_FunctionGradient(TaoLineSearch ls, Vec X, PetscReal *fcn,  Vec G, void *ptr)
81 {
82   Tao            tao = (Tao)ptr;
83   TAO_SSLS       *asls = (TAO_SSLS *)tao->data;
84 
85   PetscFunctionBegin;
86   PetscCall(TaoComputeConstraints(tao, X, tao->constraints));
87   PetscCall(VecFischer(X,tao->constraints,tao->XL,tao->XU,asls->ff));
88   PetscCall(VecNorm(asls->ff,NORM_2,&asls->merit));
89   *fcn = 0.5*asls->merit*asls->merit;
90 
91   PetscCall(TaoComputeJacobian(tao,tao->solution,tao->jacobian,tao->jacobian_pre));
92   PetscCall(MatDFischer(tao->jacobian, tao->solution, tao->constraints,tao->XL, tao->XU, asls->t1, asls->t2,asls->da, asls->db));
93   PetscCall(VecPointwiseMult(asls->t1, asls->ff, asls->db));
94   PetscCall(MatMultTranspose(tao->jacobian,asls->t1,G));
95   PetscCall(VecPointwiseMult(asls->t1, asls->ff, asls->da));
96   PetscCall(VecAXPY(G,1.0,asls->t1));
97   PetscFunctionReturn(0);
98 }
99 
100 static PetscErrorCode TaoDestroy_ASILS(Tao tao)
101 {
102   TAO_SSLS       *ssls = (TAO_SSLS *)tao->data;
103 
104   PetscFunctionBegin;
105   PetscCall(VecDestroy(&ssls->ff));
106   PetscCall(VecDestroy(&ssls->dpsi));
107   PetscCall(VecDestroy(&ssls->da));
108   PetscCall(VecDestroy(&ssls->db));
109   PetscCall(VecDestroy(&ssls->w));
110   PetscCall(VecDestroy(&ssls->t1));
111   PetscCall(VecDestroy(&ssls->t2));
112   PetscCall(VecDestroy(&ssls->r1));
113   PetscCall(VecDestroy(&ssls->r2));
114   PetscCall(VecDestroy(&ssls->r3));
115   PetscCall(VecDestroy(&ssls->dxfree));
116   PetscCall(MatDestroy(&ssls->J_sub));
117   PetscCall(MatDestroy(&ssls->Jpre_sub));
118   PetscCall(ISDestroy(&ssls->fixed));
119   PetscCall(ISDestroy(&ssls->free));
120   PetscCall(KSPDestroy(&tao->ksp));
121   PetscCall(PetscFree(tao->data));
122   PetscFunctionReturn(0);
123 }
124 
125 static PetscErrorCode TaoSolve_ASILS(Tao tao)
126 {
127   TAO_SSLS                     *asls = (TAO_SSLS *)tao->data;
128   PetscReal                    psi,ndpsi, normd, innerd, t=0;
129   PetscInt                     nf;
130   TaoLineSearchConvergedReason ls_reason;
131 
132   PetscFunctionBegin;
133   /* Assume that Setup has been called!
134      Set the structure for the Jacobian and create a linear solver. */
135 
136   PetscCall(TaoComputeVariableBounds(tao));
137   PetscCall(TaoLineSearchSetObjectiveAndGradientRoutine(tao->linesearch,Tao_ASLS_FunctionGradient,tao));
138   PetscCall(TaoLineSearchSetObjectiveRoutine(tao->linesearch,Tao_SSLS_Function,tao));
139 
140   /* Calculate the function value and fischer function value at the
141      current iterate */
142   PetscCall(TaoLineSearchComputeObjectiveAndGradient(tao->linesearch,tao->solution,&psi,asls->dpsi));
143   PetscCall(VecNorm(asls->dpsi,NORM_2,&ndpsi));
144 
145   tao->reason = TAO_CONTINUE_ITERATING;
146   while (1) {
147     /* Check the termination criteria */
148     PetscCall(PetscInfo(tao,"iter %" PetscInt_FMT ", merit: %g, ||dpsi||: %g\n",tao->niter, (double)asls->merit,  (double)ndpsi));
149     PetscCall(TaoLogConvergenceHistory(tao,asls->merit,ndpsi,0.0,tao->ksp_its));
150     PetscCall(TaoMonitor(tao,tao->niter,asls->merit,ndpsi,0.0,t));
151     PetscCall((*tao->ops->convergencetest)(tao,tao->cnvP));
152     if (TAO_CONTINUE_ITERATING != tao->reason) break;
153 
154     /* Call general purpose update function */
155     if (tao->ops->update) PetscCall((*tao->ops->update)(tao, tao->niter, tao->user_update));
156     tao->niter++;
157 
158     /* We are going to solve a linear system of equations.  We need to
159        set the tolerances for the solve so that we maintain an asymptotic
160        rate of convergence that is superlinear.
161        Note: these tolerances are for the reduced system.  We really need
162        to make sure that the full system satisfies the full-space conditions.
163 
164        This rule gives superlinear asymptotic convergence
165        asls->atol = min(0.5, asls->merit*sqrt(asls->merit));
166        asls->rtol = 0.0;
167 
168        This rule gives quadratic asymptotic convergence
169        asls->atol = min(0.5, asls->merit*asls->merit);
170        asls->rtol = 0.0;
171 
172        Calculate a free and fixed set of variables.  The fixed set of
173        variables are those for the d_b is approximately equal to zero.
174        The definition of approximately changes as we approach the solution
175        to the problem.
176 
177        No one rule is guaranteed to work in all cases.  The following
178        definition is based on the norm of the Jacobian matrix.  If the
179        norm is large, the tolerance becomes smaller. */
180     PetscCall(MatNorm(tao->jacobian,NORM_1,&asls->identifier));
181     asls->identifier = PetscMin(asls->merit, 1e-2) / (1 + asls->identifier);
182 
183     PetscCall(VecSet(asls->t1,-asls->identifier));
184     PetscCall(VecSet(asls->t2, asls->identifier));
185 
186     PetscCall(ISDestroy(&asls->fixed));
187     PetscCall(ISDestroy(&asls->free));
188     PetscCall(VecWhichBetweenOrEqual(asls->t1, asls->db, asls->t2, &asls->fixed));
189     PetscCall(ISComplementVec(asls->fixed,asls->t1, &asls->free));
190 
191     PetscCall(ISGetSize(asls->fixed,&nf));
192     PetscCall(PetscInfo(tao,"Number of fixed variables: %" PetscInt_FMT "\n", nf));
193 
194     /* We now have our partition.  Now calculate the direction in the
195        fixed variable space. */
196     PetscCall(TaoVecGetSubVec(asls->ff, asls->fixed, tao->subset_type, 0.0, &asls->r1));
197     PetscCall(TaoVecGetSubVec(asls->da, asls->fixed, tao->subset_type, 1.0, &asls->r2));
198     PetscCall(VecPointwiseDivide(asls->r1,asls->r1,asls->r2));
199     PetscCall(VecSet(tao->stepdirection,0.0));
200     PetscCall(VecISAXPY(tao->stepdirection, asls->fixed,1.0,asls->r1));
201 
202     /* Our direction in the Fixed Variable Set is fixed.  Calculate the
203        information needed for the step in the Free Variable Set.  To
204        do this, we need to know the diagonal perturbation and the
205        right hand side. */
206 
207     PetscCall(TaoVecGetSubVec(asls->da, asls->free, tao->subset_type, 0.0, &asls->r1));
208     PetscCall(TaoVecGetSubVec(asls->ff, asls->free, tao->subset_type, 0.0, &asls->r2));
209     PetscCall(TaoVecGetSubVec(asls->db, asls->free, tao->subset_type, 1.0, &asls->r3));
210     PetscCall(VecPointwiseDivide(asls->r1,asls->r1, asls->r3));
211     PetscCall(VecPointwiseDivide(asls->r2,asls->r2, asls->r3));
212 
213     /* r1 is the diagonal perturbation
214        r2 is the right hand side
215        r3 is no longer needed
216 
217        Now need to modify r2 for our direction choice in the fixed
218        variable set:  calculate t1 = J*d, take the reduced vector
219        of t1 and modify r2. */
220 
221     PetscCall(MatMult(tao->jacobian, tao->stepdirection, asls->t1));
222     PetscCall(TaoVecGetSubVec(asls->t1,asls->free,tao->subset_type,0.0,&asls->r3));
223     PetscCall(VecAXPY(asls->r2, -1.0, asls->r3));
224 
225     /* Calculate the reduced problem matrix and the direction */
226     if (!asls->w && (tao->subset_type == TAO_SUBSET_MASK || tao->subset_type == TAO_SUBSET_MATRIXFREE)) {
227       PetscCall(VecDuplicate(tao->solution, &asls->w));
228     }
229     PetscCall(TaoMatGetSubMat(tao->jacobian, asls->free, asls->w, tao->subset_type,&asls->J_sub));
230     if (tao->jacobian != tao->jacobian_pre) {
231       PetscCall(TaoMatGetSubMat(tao->jacobian_pre, asls->free, asls->w, tao->subset_type, &asls->Jpre_sub));
232     } else {
233       PetscCall(MatDestroy(&asls->Jpre_sub));
234       asls->Jpre_sub = asls->J_sub;
235       PetscCall(PetscObjectReference((PetscObject)(asls->Jpre_sub)));
236     }
237     PetscCall(MatDiagonalSet(asls->J_sub, asls->r1,ADD_VALUES));
238     PetscCall(TaoVecGetSubVec(tao->stepdirection, asls->free, tao->subset_type, 0.0, &asls->dxfree));
239     PetscCall(VecSet(asls->dxfree, 0.0));
240 
241     /* Calculate the reduced direction.  (Really negative of Newton
242        direction.  Therefore, rest of the code uses -d.) */
243     PetscCall(KSPReset(tao->ksp));
244     PetscCall(KSPSetOperators(tao->ksp, asls->J_sub, asls->Jpre_sub));
245     PetscCall(KSPSolve(tao->ksp, asls->r2, asls->dxfree));
246     PetscCall(KSPGetIterationNumber(tao->ksp,&tao->ksp_its));
247     tao->ksp_tot_its+=tao->ksp_its;
248 
249     /* Add the direction in the free variables back into the real direction. */
250     PetscCall(VecISAXPY(tao->stepdirection, asls->free, 1.0,asls->dxfree));
251 
252     /* Check the real direction for descent and if not, use the negative
253        gradient direction. */
254     PetscCall(VecNorm(tao->stepdirection, NORM_2, &normd));
255     PetscCall(VecDot(tao->stepdirection, asls->dpsi, &innerd));
256 
257     if (innerd <= asls->delta*PetscPowReal(normd, asls->rho)) {
258       PetscCall(PetscInfo(tao,"Gradient direction: %5.4e.\n", (double)innerd));
259       PetscCall(PetscInfo(tao, "Iteration %" PetscInt_FMT ": newton direction not descent\n", tao->niter));
260       PetscCall(VecCopy(asls->dpsi, tao->stepdirection));
261       PetscCall(VecDot(asls->dpsi, tao->stepdirection, &innerd));
262     }
263 
264     PetscCall(VecScale(tao->stepdirection, -1.0));
265     innerd = -innerd;
266 
267     /* We now have a correct descent direction.  Apply a linesearch to
268        find the new iterate. */
269     PetscCall(TaoLineSearchSetInitialStepLength(tao->linesearch, 1.0));
270     PetscCall(TaoLineSearchApply(tao->linesearch, tao->solution, &psi,asls->dpsi, tao->stepdirection, &t, &ls_reason));
271     PetscCall(VecNorm(asls->dpsi, NORM_2, &ndpsi));
272   }
273   PetscFunctionReturn(0);
274 }
275 
276 /* ---------------------------------------------------------- */
277 /*MC
278    TAOASILS - Active-set infeasible linesearch algorithm for solving
279        complementarity constraints
280 
281    Options Database Keys:
282 + -tao_ssls_delta - descent test fraction
283 - -tao_ssls_rho - descent test power
284 
285   Level: beginner
286 M*/
287 PETSC_EXTERN PetscErrorCode TaoCreate_ASILS(Tao tao)
288 {
289   TAO_SSLS       *asls;
290   const char     *armijo_type = TAOLINESEARCHARMIJO;
291 
292   PetscFunctionBegin;
293   PetscCall(PetscNewLog(tao,&asls));
294   tao->data = (void*)asls;
295   tao->ops->solve = TaoSolve_ASILS;
296   tao->ops->setup = TaoSetUp_ASILS;
297   tao->ops->view = TaoView_SSLS;
298   tao->ops->setfromoptions = TaoSetFromOptions_SSLS;
299   tao->ops->destroy = TaoDestroy_ASILS;
300   tao->subset_type = TAO_SUBSET_SUBVEC;
301   asls->delta = 1e-10;
302   asls->rho = 2.1;
303   asls->fixed = NULL;
304   asls->free = NULL;
305   asls->J_sub = NULL;
306   asls->Jpre_sub = NULL;
307   asls->w = NULL;
308   asls->r1 = NULL;
309   asls->r2 = NULL;
310   asls->r3 = NULL;
311   asls->t1 = NULL;
312   asls->t2 = NULL;
313   asls->dxfree = NULL;
314 
315   asls->identifier = 1e-5;
316 
317   PetscCall(TaoLineSearchCreate(((PetscObject)tao)->comm, &tao->linesearch));
318   PetscCall(PetscObjectIncrementTabLevel((PetscObject)tao->linesearch, (PetscObject)tao, 1));
319   PetscCall(TaoLineSearchSetType(tao->linesearch, armijo_type));
320   PetscCall(TaoLineSearchSetOptionsPrefix(tao->linesearch,tao->hdr.prefix));
321   PetscCall(TaoLineSearchSetFromOptions(tao->linesearch));
322 
323   PetscCall(KSPCreate(((PetscObject)tao)->comm, &tao->ksp));
324   PetscCall(PetscObjectIncrementTabLevel((PetscObject)tao->ksp, (PetscObject)tao, 1));
325   PetscCall(KSPSetOptionsPrefix(tao->ksp,tao->hdr.prefix));
326   PetscCall(KSPSetFromOptions(tao->ksp));
327 
328   /* Override default settings (unless already changed) */
329   if (!tao->max_it_changed) tao->max_it = 2000;
330   if (!tao->max_funcs_changed) tao->max_funcs = 4000;
331   if (!tao->gttol_changed) tao->gttol = 0;
332   if (!tao->grtol_changed) tao->grtol = 0;
333 #if defined(PETSC_USE_REAL_SINGLE)
334   if (!tao->gatol_changed) tao->gatol = 1.0e-6;
335   if (!tao->fmin_changed)  tao->fmin = 1.0e-4;
336 #else
337   if (!tao->gatol_changed) tao->gatol = 1.0e-16;
338   if (!tao->fmin_changed) tao->fmin = 1.0e-8;
339 #endif
340   PetscFunctionReturn(0);
341 }
342