xref: /petsc/src/tao/unconstrained/tutorials/rosenbrock2.c (revision bd35522dd00d2ebcbc9f656e2902fae240dc8904)
1 /* Program usage: mpiexec -n 1 rosenbrock2 [-help] [all TAO options] */
2 
3 /*  Include "petsctao.h" so we can use TAO solvers.  */
4 #include <petsctao.h>
5 
6 static  char help[] = "This example demonstrates use of the TAO package to \n\
7 solve an unconstrained minimization problem on a single processor.  We \n\
8 minimize the extended Rosenbrock function: \n\
9    sum_{i=0}^{n/2-1} (alpha*(x_{2i+1}-x_{2i}^2)^2 + (1-x_{2i})^2) \n\
10 or the chained Rosenbrock function:\n\
11    sum_{i=0}^{n-1} alpha*(x_{i+1} - x_i^2)^2 + (1 - x_i)^2\n";
12 
13 /*
14    User-defined application context - contains data needed by the
15    application-provided call-back routines that evaluate the function,
16    gradient, and hessian.
17 */
18 typedef struct {
19   PetscInt  n;          /* dimension */
20   PetscReal alpha;   /* condition parameter */
21   PetscBool chained;
22 } AppCtx;
23 
24 /* -------------- User-defined routines ---------- */
25 PetscErrorCode FormFunctionGradient(Tao,Vec,PetscReal*,Vec,void*);
26 PetscErrorCode FormHessian(Tao,Vec,Mat,Mat,void*);
27 
28 int main(int argc,char **argv)
29 {
30   PetscReal          zero=0.0;
31   Vec                x;                     /* solution vector */
32   Mat                H;
33   Tao                tao;                   /* Tao solver context */
34   PetscBool          flg, test_lmvm = PETSC_FALSE;
35   PetscMPIInt        size;                  /* number of processes running */
36   AppCtx             user;                  /* user-defined application context */
37   TaoConvergedReason reason;
38   PetscInt           its, recycled_its=0, oneshot_its=0;
39 
40   /* Initialize TAO and PETSc */
41   PetscFunctionBeginUser;
42   PetscCall(PetscInitialize(&argc,&argv,(char*)0,help));
43   PetscCallMPI(MPI_Comm_size(PETSC_COMM_WORLD,&size));
44   PetscCheck(size == 1,PETSC_COMM_WORLD,PETSC_ERR_WRONG_MPI_SIZE,"Incorrect number of processors");
45 
46   /* Initialize problem parameters */
47   user.n = 2; user.alpha = 99.0; user.chained = PETSC_FALSE;
48   /* Check for command line arguments to override defaults */
49   PetscCall(PetscOptionsGetInt(NULL,NULL,"-n",&user.n,&flg));
50   PetscCall(PetscOptionsGetReal(NULL,NULL,"-alpha",&user.alpha,&flg));
51   PetscCall(PetscOptionsGetBool(NULL,NULL,"-chained",&user.chained,&flg));
52   PetscCall(PetscOptionsGetBool(NULL,NULL,"-test_lmvm",&test_lmvm,&flg));
53 
54   /* Allocate vectors for the solution and gradient */
55   PetscCall(VecCreateSeq(PETSC_COMM_SELF,user.n,&x));
56   PetscCall(MatCreateSeqBAIJ(PETSC_COMM_SELF,2,user.n,user.n,1,NULL,&H));
57 
58   /* The TAO code begins here */
59 
60   /* Create TAO solver with desired solution method */
61   PetscCall(TaoCreate(PETSC_COMM_SELF,&tao));
62   PetscCall(TaoSetType(tao,TAOLMVM));
63 
64   /* Set solution vec and an initial guess */
65   PetscCall(VecSet(x, zero));
66   PetscCall(TaoSetSolution(tao,x));
67 
68   /* Set routines for function, gradient, hessian evaluation */
69   PetscCall(TaoSetObjectiveAndGradient(tao,NULL,FormFunctionGradient,&user));
70   PetscCall(TaoSetHessian(tao,H,H,FormHessian,&user));
71 
72   /* Check for TAO command line options */
73   PetscCall(TaoSetFromOptions(tao));
74 
75   /* Solve the problem */
76   PetscCall(TaoSetTolerances(tao, 1.e-5, 0.0, 0.0));
77   PetscCall(TaoSetMaximumIterations(tao, 5));
78   PetscCall(TaoLMVMRecycle(tao, PETSC_TRUE));
79   reason = TAO_CONTINUE_ITERATING;
80   while (reason != TAO_CONVERGED_GATOL) {
81     PetscCall(TaoSolve(tao));
82     PetscCall(TaoGetConvergedReason(tao, &reason));
83     PetscCall(TaoGetIterationNumber(tao, &its));
84     recycled_its += its;
85     PetscCall(PetscPrintf(PETSC_COMM_SELF, "-----------------------\n"));
86   }
87 
88   /* Disable recycling and solve again! */
89   PetscCall(TaoSetMaximumIterations(tao, 100));
90   PetscCall(TaoLMVMRecycle(tao, PETSC_FALSE));
91   PetscCall(VecSet(x, zero));
92   PetscCall(TaoSolve(tao));
93   PetscCall(TaoGetConvergedReason(tao, &reason));
94   PetscCheck(reason == TAO_CONVERGED_GATOL,PETSC_COMM_SELF, PETSC_ERR_NOT_CONVERGED, "Solution failed to converge!");
95   PetscCall(TaoGetIterationNumber(tao, &oneshot_its));
96   PetscCall(PetscPrintf(PETSC_COMM_SELF, "-----------------------\n"));
97   PetscCall(PetscPrintf(PETSC_COMM_SELF, "recycled its: %" PetscInt_FMT " | oneshot its: %" PetscInt_FMT "\n", recycled_its, oneshot_its));
98   PetscCheck(recycled_its == oneshot_its,PETSC_COMM_SELF, PETSC_ERR_NOT_CONVERGED, "LMVM recycling does not work!");
99 
100   PetscCall(TaoDestroy(&tao));
101   PetscCall(VecDestroy(&x));
102   PetscCall(MatDestroy(&H));
103 
104   PetscCall(PetscFinalize());
105   return 0;
106 }
107 
108 /* -------------------------------------------------------------------- */
109 /*
110     FormFunctionGradient - Evaluates the function, f(X), and gradient, G(X).
111 
112     Input Parameters:
113 .   tao  - the Tao context
114 .   X    - input vector
115 .   ptr  - optional user-defined context, as set by TaoSetFunctionGradient()
116 
117     Output Parameters:
118 .   G - vector containing the newly evaluated gradient
119 .   f - function value
120 
121     Note:
122     Some optimization methods ask for the function and the gradient evaluation
123     at the same time.  Evaluating both at once may be more efficient that
124     evaluating each separately.
125 */
126 PetscErrorCode FormFunctionGradient(Tao tao,Vec X,PetscReal *f, Vec G,void *ptr)
127 {
128   AppCtx            *user = (AppCtx *) ptr;
129   PetscInt          i,nn=user->n/2;
130   PetscReal         ff=0,t1,t2,alpha=user->alpha;
131   PetscScalar       *g;
132   const PetscScalar *x;
133 
134   PetscFunctionBeginUser;
135   /* Get pointers to vector data */
136   PetscCall(VecGetArrayRead(X,&x));
137   PetscCall(VecGetArray(G,&g));
138 
139   /* Compute G(X) */
140   if (user->chained) {
141     g[0] = 0;
142     for (i=0; i<user->n-1; i++) {
143       t1 = x[i+1] - x[i]*x[i];
144       ff += PetscSqr(1 - x[i]) + alpha*t1*t1;
145       g[i] += -2*(1 - x[i]) + 2*alpha*t1*(-2*x[i]);
146       g[i+1] = 2*alpha*t1;
147     }
148   } else {
149     for (i=0; i<nn; i++) {
150       t1 = x[2*i+1]-x[2*i]*x[2*i]; t2= 1-x[2*i];
151       ff += alpha*t1*t1 + t2*t2;
152       g[2*i] = -4*alpha*t1*x[2*i]-2.0*t2;
153       g[2*i+1] = 2*alpha*t1;
154     }
155   }
156 
157   /* Restore vectors */
158   PetscCall(VecRestoreArrayRead(X,&x));
159   PetscCall(VecRestoreArray(G,&g));
160   *f   = ff;
161 
162   PetscCall(PetscLogFlops(15.0*nn));
163   PetscFunctionReturn(0);
164 }
165 
166 /* ------------------------------------------------------------------- */
167 /*
168    FormHessian - Evaluates Hessian matrix.
169 
170    Input Parameters:
171 .  tao   - the Tao context
172 .  x     - input vector
173 .  ptr   - optional user-defined context, as set by TaoSetHessian()
174 
175    Output Parameters:
176 .  H     - Hessian matrix
177 
178    Note:  Providing the Hessian may not be necessary.  Only some solvers
179    require this matrix.
180 */
181 PetscErrorCode FormHessian(Tao tao,Vec X,Mat H, Mat Hpre, void *ptr)
182 {
183   AppCtx            *user = (AppCtx*)ptr;
184   PetscInt          i, ind[2];
185   PetscReal         alpha=user->alpha;
186   PetscReal         v[2][2];
187   const PetscScalar *x;
188   PetscBool         assembled;
189 
190   PetscFunctionBeginUser;
191   /* Zero existing matrix entries */
192   PetscCall(MatAssembled(H,&assembled));
193   if (assembled) PetscCall(MatZeroEntries(H));
194 
195   /* Get a pointer to vector data */
196   PetscCall(VecGetArrayRead(X,&x));
197 
198   /* Compute H(X) entries */
199   if (user->chained) {
200     PetscCall(MatZeroEntries(H));
201     for (i=0; i<user->n-1; i++) {
202       PetscScalar t1 = x[i+1] - x[i]*x[i];
203       v[0][0] = 2 + 2*alpha*(t1*(-2) - 2*x[i]);
204       v[0][1] = 2*alpha*(-2*x[i]);
205       v[1][0] = 2*alpha*(-2*x[i]);
206       v[1][1] = 2*alpha*t1;
207       ind[0] = i; ind[1] = i+1;
208       PetscCall(MatSetValues(H,2,ind,2,ind,v[0],ADD_VALUES));
209     }
210   } else {
211     for (i=0; i<user->n/2; i++) {
212       v[1][1] = 2*alpha;
213       v[0][0] = -4*alpha*(x[2*i+1]-3*x[2*i]*x[2*i]) + 2;
214       v[1][0] = v[0][1] = -4.0*alpha*x[2*i];
215       ind[0]=2*i; ind[1]=2*i+1;
216       PetscCall(MatSetValues(H,2,ind,2,ind,v[0],INSERT_VALUES));
217     }
218   }
219   PetscCall(VecRestoreArrayRead(X,&x));
220 
221   /* Assemble matrix */
222   PetscCall(MatAssemblyBegin(H,MAT_FINAL_ASSEMBLY));
223   PetscCall(MatAssemblyEnd(H,MAT_FINAL_ASSEMBLY));
224   PetscCall(PetscLogFlops(9.0*user->n/2.0));
225   PetscFunctionReturn(0);
226 }
227 
228 /*TEST
229 
230    build:
231       requires: !complex
232 
233    test:
234       args: -tao_type lmvm -tao_monitor
235       requires: !single
236 
237 TEST*/
238