bool
LogisticRegression::solve(double tolerance)
{
if (numObservations() <= 0)
{
THEA_WARNING << "LogisticRegression: Solving empty problem";
has_solution = true;
solution.resize(0);
}
else
{
StdLinearSolver llsq(StdLinearSolver::Method::DEFAULT, StdLinearSolver::Constraint::UNCONSTRAINED);
llsq.setTolerance(tolerance);
typedef Eigen::Map< MatrixX<double, MatrixLayout::ROW_MAJOR> > M;
M a(&llsq_coeffs[0], numObservations(), ndims);
has_solution = llsq.solve(MatrixWrapper<M>(&a), &llsq_consts[0]);
if (has_solution)
solution = Eigen::Map<VectorXd>(const_cast<double *>(llsq.getSolution()), ndims);
else
solution.resize(0);
}
return has_solution;
}
void test01 ( )
/******************************************************************************/
/*
Purpose:
TEST01 calls LLSQ to match 15 data values.
Licensing:
This code is distributed under the GNU LGPL license.
Modified:
07 March 2012
Author:
John Burkardt
*/
{
double a;
double b;
double error;
int i;
int n = 15;
double x[15] = {
1.47, 1.50, 1.52, 1.55, 1.57, 1.60, 1.63, 1.65, 1.68, 1.70,
1.73, 1.75, 1.78, 1.80, 1.83 };
double y[15] = {
52.21, 53.12, 54.48, 55.84, 57.20, 58.57, 59.93, 61.29, 63.11, 64.47,
66.28, 68.10, 69.92, 72.19, 74.46 };
printf ( "\n" );
printf ( "TEST01\n" );
printf ( " LLSQ can compute the formula for a line of the form\n" );
printf ( " y = A * x + B\n" );
printf ( " which minimizes the RMS error to a set of N data values.\n" );
llsq ( n, x, y, &a, &b );
printf ( "\n" );
printf ( " Estimated relationship is y = %g * x + %g\n", a, b );
printf ( " Expected value is y = 61.272 * x - 39.062\n" );
printf ( "\n" );
printf ( " I X Y B+A*X |error|\n" );
printf ( "\n" );
error = 0.0;
for ( i = 0; i < n; i++ )
{
printf ( " %4d %7g %7g %7g %7g\n", i, x[i], y[i], b + a * x[i], b + a * x[i] - y[i] );
error = error + pow ( b + a * x[i] - y[i], 2 );
}
error = sqrt ( error / ( double ) n );
printf ( "\n" );
printf ( " RMS error = %g\n", error );
return;
}
void Optimizer::updateNorma(const TVReal <s, const TVComplex &vs, real &normak, real &normab)
{
TVReal as(vs.size());
real mn=std::numeric_limits<real>::max();
for(size_t i(0); i<vs.size(); i++)
{
as[i] = log(vs[i].a());
if(as[i] > std::numeric_limits<real>::max() || as[i] < -std::numeric_limits<real>::max())
{
//inf
}
else
{
if(mn>as[i]) mn = as[i];
}
}
for(size_t i(0); i<vs.size(); i++)
{
if(as[i] > std::numeric_limits<real>::max() || as[i] < -std::numeric_limits<real>::max())
{
as[i] = mn;
}
}
if(mn == std::numeric_limits<real>::max())
{
normak = 0;
normab = 0;
return;
}
llsq(<s[0], &as[0], as.size(), normak, normab);
if( _isnan(normak) || _isnan(normab) ||
normak > std::numeric_limits<real>::max() || normak < -std::numeric_limits<real>::max()||
normab > std::numeric_limits<real>::max() || normab < -std::numeric_limits<real>::max())
{
normak = 0;
normab = 0;
}
}
