1 |
#include "SteepestDescent.hpp"
|
2 |
#include "Utility.hpp"
|
3 |
void SteepestDescent::minimize(){
|
4 |
int maxIteration;
|
5 |
int nextRestIter;
|
6 |
int resetFrq;
|
7 |
int nextWriteIter;
|
8 |
int writeFrq;
|
9 |
|
10 |
if (!isSolvable()){
|
11 |
cout << "ConjugateMinimizerBase Error: This nonlinear model can not be solved by " << methodName <<endl;
|
12 |
|
13 |
exit(1);
|
14 |
}
|
15 |
|
16 |
printMinizerInfo();
|
17 |
|
18 |
resetFrq = paramSet->getResetFrq();
|
19 |
nextRestIter = resetFrq;
|
20 |
|
21 |
writeFrq = paramSet->getWriteFrq();
|
22 |
nextWriteIter = writeFrq;
|
23 |
|
24 |
direction = model->calcGrad();
|
25 |
|
26 |
maxIteration = paramSet->getMaxIteration();
|
27 |
|
28 |
for(currentIter = 0;currentIter < maxIteration; currentIter++){
|
29 |
|
30 |
// perform line search to minimize f(x + stepSize * direction) where stepSize > 0
|
31 |
lsMinimizer->minimize(direction, 0.0, 1.0);
|
32 |
|
33 |
lsStatus = lsMinimizer->getMinimizationStatus();
|
34 |
|
35 |
if(lsStatus ==MINSTATUS_ERROR){
|
36 |
minStatus = MINSTATUS_ERROR;
|
37 |
return;
|
38 |
}
|
39 |
|
40 |
prevMinX = minX;
|
41 |
minX = minX + lsMinimizer->getMinVar() * direction;
|
42 |
|
43 |
//calculate the gradient
|
44 |
|
45 |
direction = model->calcGrad(minX);
|
46 |
|
47 |
// stop if converge
|
48 |
if (checkConvergence() > 0){
|
49 |
writeOut(minX, currentIter);
|
50 |
|
51 |
minStatus = MINSTATUS_CONVERGE;
|
52 |
return;
|
53 |
}
|
54 |
|
55 |
//
|
56 |
if (currentIter == nextWriteIter){
|
57 |
nextWriteIter += writeFrq;
|
58 |
writeOut(minX, currentIter);
|
59 |
}
|
60 |
|
61 |
if (currentIter == nextResetIter){
|
62 |
reset();
|
63 |
nextResetIter += resetFrq;
|
64 |
}
|
65 |
|
66 |
}
|
67 |
|
68 |
// if writeFrq is not a multipiler of maxIteration, we need to write the final result
|
69 |
// otherwise, we already write it inside the loop, just skip it
|
70 |
if(currentIter != (nextWriteIter - writeFrq))
|
71 |
writeOut(minX, currentIter);
|
72 |
|
73 |
minStatus = MINSTATUS_MAXITER;
|
74 |
return;
|
75 |
} |