xref: /petsc/src/tao/leastsquares/tests/chwirut2.c (revision c3b5f7ba6bc5ce25a01a67bb37ba5d34b02bbbd7)
1 /*
2    Include "petsctao.h" so that we can use TAO solvers.  Note that this
3    file automatically includes libraries such as:
4      petsc.h       - base PETSc routines   petscvec.h - vectors
5      petscsys.h    - system routines        petscmat.h - matrices
6      petscis.h     - index sets            petscksp.h - Krylov subspace methods
7      petscviewer.h - viewers               petscpc.h  - preconditioners
8 
9  This version tests correlated terms using both vector and listed forms
10 */
11 
12 #include <petsctao.h>
13 
14 /*
15 Description:   These data are the result of a NIST study involving
16                ultrasonic calibration.  The response variable is
17                ultrasonic response, and the predictor variable is
18                metal distance.
19 
20 Reference:     Chwirut, D., NIST (197?).
21                Ultrasonic Reference Block Study.
22 */
23 
24 static char help[]="Finds the nonlinear least-squares solution to the model \n\
25             y = exp[-b1*x]/(b2+b3*x)  +  e \n";
26 
27 #define NOBSERVATIONS 214
28 #define NPARAMETERS 3
29 
30 /* User-defined application context */
31 typedef struct {
32   /* Working space */
33   PetscReal t[NOBSERVATIONS];   /* array of independent variables of observation */
34   PetscReal y[NOBSERVATIONS];   /* array of dependent variables */
35   PetscReal j[NOBSERVATIONS][NPARAMETERS]; /* dense jacobian matrix array*/
36   PetscInt idm[NOBSERVATIONS];  /* Matrix indices for jacobian */
37   PetscInt idn[NPARAMETERS];
38 } AppCtx;
39 
40 /* User provided Routines */
41 PetscErrorCode InitializeData(AppCtx *user);
42 PetscErrorCode FormStartingPoint(Vec);
43 PetscErrorCode EvaluateFunction(Tao, Vec, Vec, void *);
44 PetscErrorCode EvaluateJacobian(Tao, Vec, Mat, Mat, void *);
45 
46 /*--------------------------------------------------------------------*/
47 int main(int argc,char **argv)
48 {
49   PetscInt       wtype=0;
50   Vec            x, f;               /* solution, function */
51   Vec            w;                  /* weights */
52   Mat            J;                  /* Jacobian matrix */
53   Tao            tao;                /* Tao solver context */
54   PetscInt       i;               /* iteration information */
55   PetscReal      hist[100],resid[100];
56   PetscInt       lits[100];
57   PetscInt       w_row[NOBSERVATIONS]; /* explicit weights */
58   PetscInt       w_col[NOBSERVATIONS];
59   PetscReal      w_vals[NOBSERVATIONS];
60   PetscBool      flg;
61   AppCtx         user;               /* user-defined work context */
62 
63   PetscCall(PetscInitialize(&argc,&argv,(char *)0,help));
64   PetscCall(PetscOptionsGetInt(NULL,NULL,"-wtype",&wtype,&flg));
65   PetscCall(PetscPrintf(PETSC_COMM_WORLD,"wtype=%d\n",wtype));
66   /* Allocate vectors */
67   PetscCall(VecCreateSeq(MPI_COMM_SELF,NPARAMETERS,&x));
68   PetscCall(VecCreateSeq(MPI_COMM_SELF,NOBSERVATIONS,&f));
69 
70   PetscCall(VecDuplicate(f,&w));
71 
72   /* no correlation, but set in different ways */
73   PetscCall(VecSet(w,1.0));
74   for (i=0;i<NOBSERVATIONS;i++) {
75     w_row[i]=i; w_col[i]=i; w_vals[i]=1.0;
76   }
77 
78   /* Create the Jacobian matrix. */
79   PetscCall(MatCreateSeqDense(MPI_COMM_SELF,NOBSERVATIONS,NPARAMETERS,NULL,&J));
80 
81   for (i=0;i<NOBSERVATIONS;i++) user.idm[i] = i;
82 
83   for (i=0;i<NPARAMETERS;i++) user.idn[i] = i;
84 
85   /* Create TAO solver and set desired solution method */
86   PetscCall(TaoCreate(PETSC_COMM_SELF,&tao));
87   PetscCall(TaoSetType(tao,TAOPOUNDERS));
88 
89  /* Set the function and Jacobian routines. */
90   PetscCall(InitializeData(&user));
91   PetscCall(FormStartingPoint(x));
92   PetscCall(TaoSetSolution(tao,x));
93   PetscCall(TaoSetResidualRoutine(tao,f,EvaluateFunction,(void*)&user));
94   if (wtype == 1) {
95     PetscCall(TaoSetResidualWeights(tao,w,0,NULL,NULL,NULL));
96   } else if (wtype == 2) {
97     PetscCall(TaoSetResidualWeights(tao,NULL,NOBSERVATIONS,w_row,w_col,w_vals));
98   }
99   PetscCall(TaoSetJacobianResidualRoutine(tao, J, J, EvaluateJacobian, (void*)&user));
100   PetscCall(TaoSetTolerances(tao,1e-5,0.0,PETSC_DEFAULT));
101 
102   /* Check for any TAO command line arguments */
103   PetscCall(TaoSetFromOptions(tao));
104 
105   PetscCall(TaoSetConvergenceHistory(tao,hist,resid,0,lits,100,PETSC_TRUE));
106   /* Perform the Solve */
107   PetscCall(TaoSolve(tao));
108 
109   /* Free TAO data structures */
110   PetscCall(TaoDestroy(&tao));
111 
112    /* Free PETSc data structures */
113   PetscCall(VecDestroy(&x));
114   PetscCall(VecDestroy(&w));
115   PetscCall(VecDestroy(&f));
116   PetscCall(MatDestroy(&J));
117 
118   PetscCall(PetscFinalize());
119   return 0;
120 }
121 
122 /*--------------------------------------------------------------------*/
123 PetscErrorCode EvaluateFunction(Tao tao, Vec X, Vec F, void *ptr)
124 {
125   AppCtx          *user = (AppCtx *)ptr;
126   PetscInt        i;
127   PetscReal       *y=user->y,*f,*t=user->t;
128   const PetscReal *x;
129 
130   PetscFunctionBegin;
131   PetscCall(VecGetArrayRead(X,&x));
132   PetscCall(VecGetArray(F,&f));
133 
134   for (i=0;i<NOBSERVATIONS;i++) {
135     f[i] = y[i] - PetscExpScalar(-x[0]*t[i])/(x[1] + x[2]*t[i]);
136   }
137   PetscCall(VecRestoreArrayRead(X,&x));
138   PetscCall(VecRestoreArray(F,&f));
139   PetscLogFlops(6*NOBSERVATIONS);
140   PetscFunctionReturn(0);
141 }
142 
143 /*------------------------------------------------------------*/
144 /* J[i][j] = df[i]/dt[j] */
145 PetscErrorCode EvaluateJacobian(Tao tao, Vec X, Mat J, Mat Jpre, void *ptr)
146 {
147   AppCtx          *user = (AppCtx *)ptr;
148   PetscInt        i;
149   PetscReal       *t=user->t;
150   const PetscReal *x;
151   PetscReal       base;
152 
153   PetscFunctionBegin;
154   PetscCall(VecGetArrayRead(X,&x));
155   for (i=0;i<NOBSERVATIONS;i++) {
156     base = PetscExpScalar(-x[0]*t[i])/(x[1] + x[2]*t[i]);
157 
158     user->j[i][0] = t[i]*base;
159     user->j[i][1] = base/(x[1] + x[2]*t[i]);
160     user->j[i][2] = base*t[i]/(x[1] + x[2]*t[i]);
161   }
162 
163   /* Assemble the matrix */
164   PetscCall(MatSetValues(J,NOBSERVATIONS,user->idm, NPARAMETERS, user->idn,(PetscReal *)user->j,INSERT_VALUES));
165   PetscCall(MatAssemblyBegin(J,MAT_FINAL_ASSEMBLY));
166   PetscCall(MatAssemblyEnd(J,MAT_FINAL_ASSEMBLY));
167 
168   PetscCall(VecRestoreArrayRead(X,&x));
169   PetscLogFlops(NOBSERVATIONS * 13);
170   PetscFunctionReturn(0);
171 }
172 
173 /* ------------------------------------------------------------ */
174 PetscErrorCode FormStartingPoint(Vec X)
175 {
176   PetscReal      *x;
177 
178   PetscFunctionBegin;
179   PetscCall(VecGetArray(X,&x));
180   x[0] = 1.19;
181   x[1] = -1.86;
182   x[2] = 1.08;
183   PetscCall(VecRestoreArray(X,&x));
184   PetscFunctionReturn(0);
185 }
186 
187 /* ---------------------------------------------------------------------- */
188 PetscErrorCode InitializeData(AppCtx *user)
189 {
190   PetscReal *t=user->t,*y=user->y;
191   PetscInt  i=0;
192 
193   PetscFunctionBegin;
194   y[i] =   92.9000;   t[i++] =  0.5000;
195   y[i] =    78.7000;  t[i++] =   0.6250;
196   y[i] =    64.2000;  t[i++] =   0.7500;
197   y[i] =    64.9000;  t[i++] =   0.8750;
198   y[i] =    57.1000;  t[i++] =   1.0000;
199   y[i] =    43.3000;  t[i++] =   1.2500;
200   y[i] =    31.1000;   t[i++] =  1.7500;
201   y[i] =    23.6000;   t[i++] =  2.2500;
202   y[i] =    31.0500;   t[i++] =  1.7500;
203   y[i] =    23.7750;   t[i++] =  2.2500;
204   y[i] =    17.7375;   t[i++] =  2.7500;
205   y[i] =    13.8000;   t[i++] =  3.2500;
206   y[i] =    11.5875;   t[i++] =  3.7500;
207   y[i] =     9.4125;   t[i++] =  4.2500;
208   y[i] =     7.7250;   t[i++] =  4.7500;
209   y[i] =     7.3500;   t[i++] =  5.2500;
210   y[i] =     8.0250;   t[i++] =  5.7500;
211   y[i] =    90.6000;   t[i++] =  0.5000;
212   y[i] =    76.9000;   t[i++] =  0.6250;
213   y[i] =    71.6000;   t[i++] = 0.7500;
214   y[i] =    63.6000;   t[i++] =  0.8750;
215   y[i] =    54.0000;   t[i++] =  1.0000;
216   y[i] =    39.2000;   t[i++] =  1.2500;
217   y[i] =    29.3000;   t[i++] = 1.7500;
218   y[i] =    21.4000;   t[i++] =  2.2500;
219   y[i] =    29.1750;   t[i++] =  1.7500;
220   y[i] =    22.1250;   t[i++] =  2.2500;
221   y[i] =    17.5125;   t[i++] =  2.7500;
222   y[i] =    14.2500;   t[i++] =  3.2500;
223   y[i] =     9.4500;   t[i++] =  3.7500;
224   y[i] =     9.1500;   t[i++] =  4.2500;
225   y[i] =     7.9125;   t[i++] =  4.7500;
226   y[i] =     8.4750;   t[i++] =  5.2500;
227   y[i] =     6.1125;   t[i++] =  5.7500;
228   y[i] =    80.0000;   t[i++] =  0.5000;
229   y[i] =    79.0000;   t[i++] =  0.6250;
230   y[i] =    63.8000;   t[i++] =  0.7500;
231   y[i] =    57.2000;   t[i++] =  0.8750;
232   y[i] =    53.2000;   t[i++] =  1.0000;
233   y[i] =   42.5000;   t[i++] =  1.2500;
234   y[i] =   26.8000;   t[i++] =  1.7500;
235   y[i] =    20.4000;   t[i++] =  2.2500;
236   y[i] =    26.8500;  t[i++] =   1.7500;
237   y[i] =    21.0000;  t[i++] =   2.2500;
238   y[i] =    16.4625;  t[i++] =   2.7500;
239   y[i] =    12.5250;  t[i++] =   3.2500;
240   y[i] =    10.5375;  t[i++] =   3.7500;
241   y[i] =     8.5875;  t[i++] =   4.2500;
242   y[i] =     7.1250;  t[i++] =   4.7500;
243   y[i] =     6.1125;  t[i++] =   5.2500;
244   y[i] =     5.9625;  t[i++] =   5.7500;
245   y[i] =    74.1000;  t[i++] =   0.5000;
246   y[i] =    67.3000;  t[i++] =   0.6250;
247   y[i] =    60.8000;  t[i++] =   0.7500;
248   y[i] =    55.5000;  t[i++] =   0.8750;
249   y[i] =    50.3000;  t[i++] =   1.0000;
250   y[i] =    41.0000;  t[i++] =   1.2500;
251   y[i] =    29.4000;  t[i++] =   1.7500;
252   y[i] =    20.4000;  t[i++] =   2.2500;
253   y[i] =    29.3625;  t[i++] =   1.7500;
254   y[i] =    21.1500;  t[i++] =   2.2500;
255   y[i] =    16.7625;  t[i++] =   2.7500;
256   y[i] =    13.2000;  t[i++] =   3.2500;
257   y[i] =    10.8750;  t[i++] =   3.7500;
258   y[i] =     8.1750;  t[i++] =   4.2500;
259   y[i] =     7.3500;  t[i++] =   4.7500;
260   y[i] =     5.9625;  t[i++] =  5.2500;
261   y[i] =     5.6250;  t[i++] =   5.7500;
262   y[i] =    81.5000;  t[i++] =    .5000;
263   y[i] =    62.4000;  t[i++] =    .7500;
264   y[i] =    32.5000;  t[i++] =   1.5000;
265   y[i] =    12.4100;  t[i++] =   3.0000;
266   y[i] =    13.1200;  t[i++] =   3.0000;
267   y[i] =    15.5600;  t[i++] =   3.0000;
268   y[i] =     5.6300;  t[i++] =   6.0000;
269   y[i] =    78.0000;   t[i++] =   .5000;
270   y[i] =    59.9000;  t[i++] =    .7500;
271   y[i] =    33.2000;  t[i++] =   1.5000;
272   y[i] =    13.8400;  t[i++] =   3.0000;
273   y[i] =    12.7500;  t[i++] =   3.0000;
274   y[i] =    14.6200;  t[i++] =   3.0000;
275   y[i] =     3.9400;  t[i++] =   6.0000;
276   y[i] =    76.8000;  t[i++] =    .5000;
277   y[i] =    61.0000;  t[i++] =    .7500;
278   y[i] =    32.9000;  t[i++] =   1.5000;
279   y[i] =   13.8700;   t[i++] = 3.0000;
280   y[i] =    11.8100;  t[i++] =   3.0000;
281   y[i] =    13.3100;  t[i++] =   3.0000;
282   y[i] =     5.4400;  t[i++] =   6.0000;
283   y[i] =    78.0000;  t[i++] =    .5000;
284   y[i] =    63.5000;  t[i++] =    .7500;
285   y[i] =    33.8000;  t[i++] =   1.5000;
286   y[i] =    12.5600;  t[i++] =   3.0000;
287   y[i] =     5.6300;  t[i++] =   6.0000;
288   y[i] =    12.7500;  t[i++] =   3.0000;
289   y[i] =    13.1200;  t[i++] =   3.0000;
290   y[i] =     5.4400;  t[i++] =   6.0000;
291   y[i] =    76.8000;  t[i++] =    .5000;
292   y[i] =    60.0000;  t[i++] =    .7500;
293   y[i] =    47.8000;  t[i++] =   1.0000;
294   y[i] =    32.0000;  t[i++] =   1.5000;
295   y[i] =    22.2000;  t[i++] =   2.0000;
296   y[i] =    22.5700;  t[i++] =   2.0000;
297   y[i] =    18.8200;  t[i++] =   2.5000;
298   y[i] =    13.9500;  t[i++] =   3.0000;
299   y[i] =    11.2500;  t[i++] =   4.0000;
300   y[i] =     9.0000;  t[i++] =   5.0000;
301   y[i] =     6.6700;  t[i++] =   6.0000;
302   y[i] =    75.8000;  t[i++] =    .5000;
303   y[i] =    62.0000;  t[i++] =    .7500;
304   y[i] =    48.8000;  t[i++] =   1.0000;
305   y[i] =    35.2000;  t[i++] =   1.5000;
306   y[i] =    20.0000;  t[i++] =   2.0000;
307   y[i] =    20.3200;  t[i++] =   2.0000;
308   y[i] =    19.3100;  t[i++] =   2.5000;
309   y[i] =    12.7500;  t[i++] =   3.0000;
310   y[i] =    10.4200;  t[i++] =   4.0000;
311   y[i] =     7.3100;  t[i++] =   5.0000;
312   y[i] =     7.4200;  t[i++] =   6.0000;
313   y[i] =    70.5000;  t[i++] =    .5000;
314   y[i] =    59.5000;  t[i++] =    .7500;
315   y[i] =    48.5000;  t[i++] =   1.0000;
316   y[i] =    35.8000;  t[i++] =   1.5000;
317   y[i] =    21.0000;  t[i++] =   2.0000;
318   y[i] =    21.6700;  t[i++] =   2.0000;
319   y[i] =    21.0000;  t[i++] =   2.5000;
320   y[i] =    15.6400;  t[i++] =   3.0000;
321   y[i] =     8.1700;  t[i++] =   4.0000;
322   y[i] =     8.5500;  t[i++] =   5.0000;
323   y[i] =    10.1200;  t[i++] =   6.0000;
324   y[i] =    78.0000;  t[i++] =    .5000;
325   y[i] =    66.0000;  t[i++] =    .6250;
326   y[i] =    62.0000;  t[i++] =    .7500;
327   y[i] =    58.0000;  t[i++] =    .8750;
328   y[i] =    47.7000;  t[i++] =   1.0000;
329   y[i] =    37.8000;  t[i++] =   1.2500;
330   y[i] =    20.2000;  t[i++] =   2.2500;
331   y[i] =    21.0700;  t[i++] =   2.2500;
332   y[i] =    13.8700;  t[i++] =   2.7500;
333   y[i] =     9.6700;  t[i++] =   3.2500;
334   y[i] =     7.7600;  t[i++] =   3.7500;
335   y[i] =    5.4400;   t[i++] =  4.2500;
336   y[i] =    4.8700;   t[i++] =  4.7500;
337   y[i] =     4.0100;  t[i++] =   5.2500;
338   y[i] =     3.7500;  t[i++] =   5.7500;
339   y[i] =    24.1900;  t[i++] =   3.0000;
340   y[i] =    25.7600;  t[i++] =   3.0000;
341   y[i] =    18.0700;  t[i++] =   3.0000;
342   y[i] =    11.8100;  t[i++] =   3.0000;
343   y[i] =    12.0700;  t[i++] =   3.0000;
344   y[i] =    16.1200;  t[i++] =   3.0000;
345   y[i] =    70.8000;  t[i++] =    .5000;
346   y[i] =    54.7000;  t[i++] =    .7500;
347   y[i] =    48.0000;  t[i++] =   1.0000;
348   y[i] =    39.8000;  t[i++] =   1.5000;
349   y[i] =    29.8000;  t[i++] =   2.0000;
350   y[i] =    23.7000;  t[i++] =   2.5000;
351   y[i] =    29.6200;  t[i++] =   2.0000;
352   y[i] =    23.8100;  t[i++] =   2.5000;
353   y[i] =    17.7000;  t[i++] =   3.0000;
354   y[i] =    11.5500;  t[i++] =   4.0000;
355   y[i] =    12.0700;  t[i++] =   5.0000;
356   y[i] =     8.7400;  t[i++] =   6.0000;
357   y[i] =    80.7000;  t[i++] =    .5000;
358   y[i] =    61.3000;  t[i++] =    .7500;
359   y[i] =    47.5000;  t[i++] =   1.0000;
360    y[i] =   29.0000;  t[i++] =   1.5000;
361    y[i] =   24.0000;  t[i++] =   2.0000;
362   y[i] =    17.7000;  t[i++] =   2.5000;
363   y[i] =    24.5600;  t[i++] =   2.0000;
364   y[i] =    18.6700;  t[i++] =   2.5000;
365    y[i] =   16.2400;  t[i++] =   3.0000;
366   y[i] =     8.7400;  t[i++] =   4.0000;
367   y[i] =     7.8700;  t[i++] =   5.0000;
368   y[i] =     8.5100;  t[i++] =   6.0000;
369   y[i] =    66.7000;  t[i++] =    .5000;
370   y[i] =    59.2000;  t[i++] =    .7500;
371   y[i] =    40.8000;  t[i++] =   1.0000;
372   y[i] =    30.7000;  t[i++] =   1.5000;
373   y[i] =    25.7000;  t[i++] =   2.0000;
374   y[i] =    16.3000;  t[i++] =   2.5000;
375   y[i] =    25.9900;  t[i++] =   2.0000;
376   y[i] =    16.9500;  t[i++] =   2.5000;
377   y[i] =    13.3500;  t[i++] =   3.0000;
378   y[i] =     8.6200;  t[i++] =   4.0000;
379   y[i] =     7.2000;  t[i++] =   5.0000;
380   y[i] =     6.6400;  t[i++] =   6.0000;
381   y[i] =    13.6900;  t[i++] =   3.0000;
382   y[i] =    81.0000;  t[i++] =    .5000;
383   y[i] =    64.5000;  t[i++] =    .7500;
384   y[i] =    35.5000;  t[i++] =   1.5000;
385    y[i] =   13.3100;  t[i++] =   3.0000;
386   y[i] =     4.8700;  t[i++] =   6.0000;
387   y[i] =    12.9400;  t[i++] =   3.0000;
388   y[i] =     5.0600;  t[i++] =   6.0000;
389   y[i] =    15.1900;  t[i++] =   3.0000;
390   y[i] =    14.6200;  t[i++] =   3.0000;
391   y[i] =    15.6400;  t[i++] =   3.0000;
392   y[i] =    25.5000;  t[i++] =   1.7500;
393   y[i] =    25.9500;  t[i++] =   1.7500;
394   y[i] =    81.7000;  t[i++] =    .5000;
395   y[i] =    61.6000;  t[i++] =    .7500;
396   y[i] =    29.8000;  t[i++] =   1.7500;
397   y[i] =    29.8100;  t[i++] =   1.7500;
398   y[i] =    17.1700;  t[i++] =   2.7500;
399   y[i] =    10.3900;  t[i++] =   3.7500;
400   y[i] =    28.4000;  t[i++] =   1.7500;
401   y[i] =    28.6900;  t[i++] =   1.7500;
402   y[i] =    81.3000;  t[i++] =    .5000;
403   y[i] =    60.9000;  t[i++] =    .7500;
404   y[i] =    16.6500;  t[i++] =   2.7500;
405   y[i] =    10.0500;  t[i++] =   3.7500;
406   y[i] =    28.9000;  t[i++] =   1.7500;
407   y[i] =    28.9500;  t[i++] =   1.7500;
408   PetscFunctionReturn(0);
409 }
410 
411 /*TEST
412 
413      build:
414        requires: !complex
415 
416      test:
417        args:  -tao_smonitor -tao_max_it 100 -tao_type pounders -tao_pounders_delta 0.05 -tao_gatol 1.e-5
418        requires: !single
419        TODO: produces different output for many different systems
420 
421      test:
422        suffix: 2
423        args: -tao_smonitor -tao_max_it 100 -wtype 1 -tao_type pounders -tao_pounders_delta 0.05 -tao_gatol 1.e-5
424        requires: !single
425        TODO: produces different output for many different systems
426 
427      test:
428        suffix: 3
429        args: -tao_smonitor -tao_max_it 100 -wtype 2 -tao_type pounders -tao_pounders_delta 0.05 -tao_gatol 1.e-5
430        requires: !single
431        TODO: produces different output for many different systems
432 
433      test:
434        suffix: 4
435        args: -tao_smonitor -tao_max_it 100 -tao_type pounders -tao_pounders_delta 0.05 -pounders_subsolver_tao_type blmvm -tao_gatol 1.e-5
436        requires: !single
437        TODO: produces different output for many different systems
438 
439  TEST*/
440