Refined interface for Conjugate Gradient

Some interface was refined (most notably, the method for returning
Hessian was removed and the method for getting gradient was added as
optional to base Solver::Function class) and basic code for
setters/getters was added. Now is the time for the real work on an
algorithm.
This commit is contained in:
Alex Leontiev 2013-09-22 00:14:49 +08:00
parent eb1333d0a8
commit 581d454536
5 changed files with 157 additions and 12 deletions

View File

@ -0,0 +1,11 @@
Conjugate Gradient
=======================
.. highlight:: cpp
optim::ConjGradSolver
---------------------------------
.. ocv:class:: optim::ConjGradSolver
This class is used

View File

@ -55,6 +55,7 @@ public:
public:
virtual ~Function() {}
virtual double calc(const double* x) const = 0;
virtual void getGradient(const double* /*x*/,double* /*grad*/) {}
};
virtual Ptr<Function> getFunction() const = 0;
@ -86,17 +87,7 @@ CV_EXPORTS_W Ptr<DownhillSolver> createDownhillSolver(const Ptr<Solver::Function
TermCriteria termcrit=TermCriteria(TermCriteria::MAX_ITER+TermCriteria::EPS,5000,0.000001));
//! conjugate gradient method
class CV_EXPORTS ConjGradSolver : public Solver
{
public:
class CV_EXPORTS Function : public Solver::Function
{
public:
//! gradient is like the first derivative for multivar function
virtual void getGradient(const double* x,double* buf)const=0;
//! Jacobian is like the second derivative
virtual void getJacobian(const double* x)const=0;
};
class CV_EXPORTS ConjGradSolver : public Solver{
};
CV_EXPORTS_W Ptr<ConjGradSolver> createConjGradSolver(const Ptr<Solver::Function>& f=Ptr<ConjGradSolver::Function>(),

View File

@ -0,0 +1,77 @@
#include "precomp.hpp"
#include "debug.hpp"
namespace cv{namespace optim{
class ConjGradSolverImpl : public ConjGradSolver
{
public:
Ptr<Function> getFunction() const;
void setFunction(const Ptr<Function>& f);
TermCriteria getTermCriteria() const;
ConjGradSolverImpl();
void setTermCriteria(const TermCriteria& termcrit);
double minimize(InputOutputArray x);
protected:
Ptr<Solver::Function> _Function;
TermCriteria _termcrit;
Mat_<double> d,r,buf_x,r_old;
private:
};
double ConjGradSolverImpl::minimize(InputOutputArray x){
CV_Assert(_Function.empty()==false);
dprintf(("termcrit:\n\ttype: %d\n\tmaxCount: %d\n\tEPS: %g\n",_termcrit.type,_termcrit.maxCount,_termcrit.epsilon));
Mat x_mat=x.getMat();
CV_Assert(MIN(x_mat.rows,x_mat.cols)==1);
int ndim=MAX(x_mat.rows,x_mat.cols);
CV_Assert(x_mat.type()==CV_64FC1);
d.create(1,ndim);
r.create(1,ndim);
r_old.create(1,ndim);
Mat_<double> proxy_x;
if(x_mat.rows>1){
buf_x.create(1,ndim);
Mat_<double> proxy(ndim,1,(double*)buf_x.data);
x_mat.copyTo(proxy);
proxy_x=buf_x;
}else{
proxy_x=x_mat;
}
//here everything goes. check that everything is setted properly
if(x_mat.rows>1){
Mat(ndim, 1, CV_64F, (double*)proxy_x.data).copyTo(x);
}
return 0.0;
}
ConjGradSolverImpl::ConjGradSolverImpl(){
_Function=Ptr<Function>();
}
Ptr<Solver::Function> ConjGradSolverImpl::getFunction()const{
return _Function;
}
void ConjGradSolverImpl::setFunction(const Ptr<Function>& f){
_Function=f;
}
TermCriteria ConjGradSolverImpl::getTermCriteria()const{
return _termcrit;
}
void ConjGradSolverImpl::setTermCriteria(const TermCriteria& termcrit){
CV_Assert((termcrit.type==(TermCriteria::MAX_ITER+TermCriteria::EPS) && termcrit.epsilon>0 && termcrit.maxCount>0) ||
((termcrit.type==TermCriteria::MAX_ITER) && termcrit.maxCount>0));
_termcrit=termcrit;
}
// both minRange & minError are specified by termcrit.epsilon; In addition, user may specify the number of iterations that the algorithm does.
Ptr<ConjGradSolver> createConjGradSolver(const Ptr<Solver::Function>& f, TermCriteria termcrit){
ConjGradSolver *CG=new ConjGradSolverImpl();
CG->setFunction(f);
CG->setTermCriteria(termcrit);
return Ptr<ConjGradSolver>(CG);
}
}}

View File

@ -19,6 +19,8 @@ namespace cv{namespace optim{
Ptr<Solver::Function> _Function;
TermCriteria _termcrit;
Mat _step;
Mat_<double> buf_x;
private:
inline void createInitialSimplex(Mat_<double>& simplex,Mat& step);
inline double innerDownhillSimplex(cv::Mat_<double>& p,double MinRange,double MinError,int& nfunk,
@ -209,7 +211,10 @@ namespace cv{namespace optim{
Mat_<double> proxy_x;
if(x_mat.rows>1){
proxy_x=x_mat.t();
buf_x.create(1,_step.cols);
Mat_<double> proxy(_step.cols,1,(double*)buf_x.data);
x_mat.copyTo(proxy);
proxy_x=buf_x;
}else{
proxy_x=x_mat;
}

View File

@ -0,0 +1,61 @@
#include "test_precomp.hpp"
#include <cstdlib>
static void mytest(cv::Ptr<cv::optim::ConjGradSolver> solver,cv::Ptr<cv::optim::Solver::Function> ptr_F,cv::Mat& x,
cv::Mat& etalon_x,double etalon_res){
solver->setFunction(ptr_F);
//int ndim=MAX(step.cols,step.rows);
double res=solver->minimize(x);
std::cout<<"res:\n\t"<<res<<std::endl;
std::cout<<"x:\n\t"<<x<<std::endl;
std::cout<<"etalon_res:\n\t"<<etalon_res<<std::endl;
std::cout<<"etalon_x:\n\t"<<etalon_x<<std::endl;
double tol=solver->getTermCriteria().epsilon;
ASSERT_TRUE(std::abs(res-etalon_res)<tol);
/*for(cv::Mat_<double>::iterator it1=x.begin<double>(),it2=etalon_x.begin<double>();it1!=x.end<double>();it1++,it2++){
ASSERT_TRUE(std::abs((*it1)-(*it2))<tol);
}*/
std::cout<<"--------------------------\n";
}
class SphereF:public cv::optim::Solver::Function{
public:
double calc(const double* x)const{
return x[0]*x[0]+x[1]*x[1]+x[2]*x[2]+x[3]*x[3];
}
void getGradient(const double* x,double* grad){
for(int i=0;i<4;i++,grad++,x++){
grad[0]=2*x[0];
}
}
};
//TODO: test transp/usual x
/*class RosenbrockF:public cv::optim::Solver::Function{
double calc(const double* x)const{
return 100*(x[1]-x[0]*x[0])*(x[1]-x[0]*x[0])+(1-x[0])*(1-x[0]);
}
};*/
TEST(Optim_ConjGrad, regression_basic){
cv::Ptr<cv::optim::ConjGradSolver> solver=cv::optim::createConjGradSolver();
#if 1
{
cv::Ptr<cv::optim::Solver::Function> ptr_F(new SphereF());
cv::Mat x=(cv::Mat_<double>(1,2)<<1.0,1.0),
etalon_x=(cv::Mat_<double>(1,2)<<0.0,0.0);
double etalon_res=0.0;
return;
mytest(solver,ptr_F,x,etalon_x,etalon_res);
}
#endif
#if 0
{
cv::Ptr<cv::optim::Solver::Function> ptr_F(new RosenbrockF());
cv::Mat x=(cv::Mat_<double>(2,1)<<0.0,0.0),
step=(cv::Mat_<double>(2,1)<<0.5,+0.5),
etalon_x=(cv::Mat_<double>(2,1)<<1.0,1.0);
double etalon_res=0.0;
mytest(solver,ptr_F,x,step,etalon_x,etalon_res);
}
#endif
}