search

seacode / gmacs

A generic size-structured stock assessment model
https://seacode.github.io/gmacs
18 stars 14 forks source link

starting with converged parameter file as input fails #141

Open jimianelli opened 9 years ago

jimianelli commented 9 years ago 
gmacs -binp gmacs.bar -phase 22

Should give the same objective function and parameter estimates as from the previously converged run....

quantifish commented 9 years ago

I just ran and got:

nloglike
 866.2094 901.6767 -49.8458 -78.9160
 4780.4873 -58.3731
 673.5021 1363.6510 1241.5066 2578.5971 149084.2094 471.5242
 167.0157
   0.0000

and when using gmacs -binp gmacs.bar -phase 22

nloglike
 866.2848 901.5892 -49.8454 -78.9161
 4780.6382 -58.3728
 673.5001 1363.6523 1241.5164 2578.5585 149084.0834 471.5202
 167.0104
   0.0000

A few decimals different here and there but largely the same. Or should they be exactly the same?

jimianelli commented 9 years ago

thanks for checking, this shows you end up at the same place (always good) but the point I was making was that running

gmacs -nox -binp gmacs.bar -phase 22

should go straight to the hessian calc and not do much if any estimation...

On Thu, Jul 16, 2015 at 4:39 PM, Darcy Webber notifications@github.com wrote:

I just ran and got:

nloglike 866.2094 901.6767 -49.8458 -78.9160 4780.4873 -58.3731 673.5021 1363.6510 1241.5066 2578.5971 149084.2094 471.5242 167.0157 0.0000

and when using gmacs -binp gmacs.bar -phase 22

nloglike 866.2848 901.5892 -49.8454 -78.9161 4780.6382 -58.3728 673.5001 1363.6523 1241.5164 2578.5585 149084.0834 471.5202 167.0104 0.0000

A few decimals different here and there but largely the same. Or should they be exactly the same?

— Reply to this email directly or view it on GitHub https://github.com/seacode/gmacs/issues/141#issuecomment-122131440.

Jim Ianelli

quantifish commented 9 years ago

Oh I see, ok just tried again and you are correct, it does not go straight to the Hessian calc...

quantifish commented 9 years ago

However, it does run twice as fast

--Runtime: 0 hours, 0 minutes, 31 seconds
--Number of function evaluations: 918
jimianelli commented 9 years ago

so the cause has to be in some initialization of a parameter (or number of parameters) outside of "INITIALIZATION_SECTION" (input parameters from commandline (-binp gmacs.bar) take precedence over this.

On Thu, Jul 16, 2015 at 5:21 PM, Darcy Webber notifications@github.com wrote:

However, it does run twice as fast

--Runtime: 0 hours, 0 minutes, 31 seconds --Number of function evaluations: 918

— Reply to this email directly or view it on GitHub https://github.com/seacode/gmacs/issues/141#issuecomment-122135817.

Jim Ianelli

quantifish commented 9 years ago

It also does this on the develop branch right? Just wanted to check it isn't something that I've stuffed up...

jimianelli commented 9 years ago

yea, think so, will double check

On Thu, Jul 16, 2015 at 10:05 PM, Darcy Webber notifications@github.com wrote:

It also does this on the develop branch right? Just wanted to check it isn't something that I've stuffed up...

— Reply to this email directly or view it on GitHub https://github.com/seacode/gmacs/issues/141#issuecomment-122176709.

Jim Ianelli