## GARANT MACROS: opt - optimization of rasonance assignments
##
## The macro assumes that the expected peaks and coherences were generated and
## the measured peaks were loaded. Two parameters are needed: the size of the
## population to use for the optimization and the name of the macro to be used
## for saving the best assignment found so far.

parameter sizePop savemacro
var       genGap t temp maxscore i j t1 t2 oldavg increment nrParents nrLocIt

set increment= 0.5		# increment for optimization schedule
set nrParents= 30		# number of assignments to select for recombination; orig = 30
set nrLocIt  = 40000		# number of iterations for local optimization; prov. 15000
eval genGap   = $sizePop/2

				# initialize population of assignments
init optimization possibilities $sizePop

				# determine sets of equivalent peaks
determine equivalent peaks

print
write log /dev/null

				# generate random intial assignments
eval temp = $sizePop - $genGap
eval t    = $nrLocIt * 2
do i 1 $temp
   print "Initializing $i(I4)"
   reset assignment
   local optimization $t
   overwrite optimization possibility 2.0 $i $i
end do

set maxscore   = -100000.0

print
print
print "Iteration   Schedule    Score      Best      Avg       Mutations"
print "================================================================"
set i   = 0.0
set j   = 1
set oldavg = 0.0
do
   # determine extend of generation
   eval generation = mod($j-1,$sizePop) / $genGap
   eval t1 = $generation * $genGap + 1
   eval t2 = ($generation + 1) * $genGap
   if ($t2 .gt. $sizePop) set t2 = $sizePop

   # score last population
   if (mod($j,$genGap) .eq. 1) then
      score population $t1 $t2
   end if

   # temprature schedule
   if ($i .le. 1.0) then
      eval temprature = 2.0
   else if ($i .le. 2.0) then
      eval temprature = 1.0
   else if ($i .le. 8.0) then
      eval temprature = 1.0 * (1.0 - ($i - 2.0) / 6.0)
   else
      eval temprature = 0.0
   end if

   # select parents
   select optimization possibilities $nrParents 2.0 $t1 $t2

   # perform optimization
   evolutionary optimization $nrLocIt $temprature

   # write child
   eval child = mod($j+$genGap-1,$sizePop) + 1
   overwrite optimization possibility 2.0 $child $child

   # status line
   print "$j(I4)   $i(F8.2)   $score(F8.2)   $maxscore(F8.2)    $avgscore(F8.2)    $mutations(F6.2)"

   # save if best solution
   if ($score .gt. $maxscore) then
      $savemacro
      set maxscore = $score
   end if

   # increase loop counters
   if (mod($j,$genGap) .eq. 0) then
      if ($avgscore .le. $oldavg) eval i = $i + $increment
      set oldavg = $avgscore
   end if
   eval j = $j + 1
   if ($i .gt. 10) break
end do

close log