A combination of SIMUL-ANN-1 with POWELL.
simul-pow-1 [nloop] [tchn] [nanne] [nsamp] [iseed] [tiny]
This method packages together SIMUL-ANN-1 and the POWELL
routine; at the end of each of the cooling
sequences, or annealing cycles, the POWELL method is
invoked. Probably one of the best choices where one `best'
answer is to be found, but at the expense of a lot
of computer time.
Note that the parameters of SIMUL-POW-1 are those of
SIMUL-POW-1 itself (which have the same meaning as in routine SIMUL-ANN-1), plus those of POWELL.
All of the SIMUL-POW-1 parameters, with the exception of
nanne, are explained under SIMUL-ANN-1.
Parameter nanne specifies the number of annealing cycles to be
used; successive anneal cycles start from cooler temperatures and have
slower cooling. The pattern of anneals that has been chosen (the
``annealing history'') is not magic in any way--for best results a
different annealing history may be better for any particular objective
function--but the pattern chosen seems to serve well.
nloop |
integer |
256 |
16 |
4096 |
tchn |
real |
0.95 |
0.1 |
0.9999 |
nanne |
integer |
16 |
1 |
256 |
nsamp |
integer |
128 |
16 |
1024 |
iseed |
integer |
14391 |
-1.e+20 |
1.e+20 |
tiny |
real |
1.e-12 |
1.e-20 |
1.e-6 |
Parameter=nloop (integer default=256 min=16 max=4096)
Maximum number of temperatures.
Parameter=tchn (real default=0.95 min=0.1 max=0.9999)
Factor for temperature reduction.
Parameter=nsamp (integer default=128 min=16 max=1024)
Number of movements at each temperature.
Parameter=iseed (integer default=14391 min=-1.e+20 max=1.e+20)
Seed for random number generator.
Parameter=tiny (real default=1.e-12 min=1.e-20 max=1.e-6)
Smallest temperature allowed.
- sherpa
-
get_method_expr,
grid,
grid-powell,
levenberg-marquardt,
method,
monte-lm,
monte-powell,
montecarlo,
powell,
sigma-rejection,
simplex,
simul-ann-1,
simul-ann-2,
simul-pow-2,
usermethod
|