example1.f90 Source File

module example1_model
!! Model for example1.

   use iso_c_binding, only: c_ptr
   use odrpack_kinds, only: dp, one, zero
   implicit none

contains

   pure subroutine fcn( &
      n, m, q, np, ldifx, beta, xplusd, ifixb, ifixx, ideval, f, fjacb, fjacd, istop, data)
   !! User-supplied subroutine for evaluating the model.

      integer, intent(in) :: n, m, q, np, ldifx, ideval, ifixb(np), ifixx(ldifx, m)
      real(dp), intent(in) :: beta(np), xplusd(n, m)
      real(dp), intent(out) :: f(n, q), fjacb(n, np, q), fjacd(n, m, q)
      integer, intent(out) :: istop
      type(c_ptr), intent(in), value :: data

      ! Local variables
      integer :: i

      ! Check for unacceptable values for this problem
      if (beta(1) < zero) then
         istop = 1
         return
      else
         istop = 0
      end if

      ! Compute predicted values
      if (mod(ideval, 10) > 0) then
         do i = 1, ubound(f, 2)
            f(:, i) = beta(1) + beta(2)*(exp(beta(3)*xplusd(:, 1)) - one)**2
         end do
      end if

      ! Compute derivatives with respect to 'beta'
      if (mod(ideval/10, 10) > 0) then
         do i = 1, ubound(f, 2)
            fjacb(:, 1, i) = one
            fjacb(:, 2, i) = (exp(beta(3)*xplusd(:, 1)) - one)**2
            fjacb(:, 3, i) = beta(2)*2*(exp(beta(3)*xplusd(:, 1)) - one)*exp(beta(3)*xplusd(:, 1))*xplusd(:, 1)
         end do
      end if

      ! Compute derivatives with respect to 'delta'
      if (mod(ideval/100, 10) > 0) then
      do i = 1, ubound(f, 2)
         fjacd(:, 1, i) = beta(2)*2*(exp(beta(3)*xplusd(:, 1)) - one)*exp(beta(3)*xplusd(:, 1))*beta(3)
      end do
      end if

   end subroutine fcn

end module example1_model

program example1
!! Explicit ODR job, with user-supplied analytic derivatives and nondefault `ifixx`.

   use odrpack_kinds, only: dp
   use odrpack, only: odr, odrpack_model
   use example1_model, only: fcn
   implicit none

   ! Variable declarations
   type(odrpack_model) :: model
   integer :: i, iprint, j, job, lundata, lunrpt, m, n, np, q
   integer, allocatable :: ifixx(:, :)
   real(dp), allocatable :: beta(:), x(:, :), y(:, :)

   ! Set model procedure
   model%fcn => fcn

   ! Set up report files
   open (newunit=lunrpt, file='./example/report1.dat')

   ! Read problem dimensions
   open (newunit=lundata, file='./example/data1.dat')
   read (lundata, *) n, m, np, q

   ! Allocate arrays
   allocate (beta(np), x(n, m), y(n, q), ifixx(n, m))

   ! Read problem data and set nondefault value for argument 'ifixx'
   read (lundata, *) (beta(i), i=1, np)
   do i = 1, n
      read (lundata, *) (x(i, j), j=1, m), (y(i, j), j=1, q)
      if (x(i, 1) == 0.0E0_dp .or. x(i, 1) == 100.0E0_dp) then
         ifixx(i, 1) = 0
      else
         ifixx(i, 1) = 1
      end if
   end do
   close (lundata)

   ! Specify task: Explicit orthogonal distance regression
   !       With user supplied derivatives (checked)
   !       Covariance matrix constructed with recomputed derivatives
   !       Delta initialized to zero
   !       Not a restart
   ! And indicate short initial report
   !       Short iteration reports every iteration, and
   !       Long final report
   job = 00020
   iprint = 1112

   ! Compute solution
   call odr(model, n, m, q, np, beta, y, x, &
            ifixx=ifixx, &
            job=job, iprint=iprint, lunerr=lunrpt, lunrpt=lunrpt)

   close (lunrpt)

end program example1