Hi Jude,
I don't know anything about how fminbnd works, but I have ported a few
optimization problems over from Matlab to Julia. You should figure out
which solution method fminbnd uses and compare them to those found here
http://ab-initio.mit.edu/wiki/index.php/NLopt_Algorithms to make sure
you're picking the best one. In general this format will get you through
almost any port. Where you just replace LN_SBPLX with the algorithm you'd
like and you add gradient information if you can.
using NLopt
## Set Initial Parameters from the outside
z=189
## Make "Container" Function because I can only optimize over x, even
though z is still a parameter
function test_max(x,z)
x[1]^2 + z
end
# keep track of # function evaluations
count = 0
function func(x::Vector, grad::Vector)
global count +=1
println("f_$count($x)")
test_max(x[1],z)
end
opt = Opt(:LN_SBPLX,1)
min_objective!(opt, func)
(minf,minx,ret)=optimize(opt,[1])
println("got $minf at $minx after $count iterations (returned $ret)")
-Alex
On Wednesday, September 3, 2014 5:12:34 AM UTC-7, Jude wrote:
>
> Hi,
>
> I am in the process of converting my Matlab code to Julia. I see that
> there is a package called NLopt available but rather than spending time
> trying to decipher the syntax, I was wondering if anyone knows the syntax
> for doing something similar to Matlab's fminbnd, eg, I have something like
> the following in Matlab and I would like to know how to do it in Julia:
>
> [x, value] = fminbnd(@(x) objectivefunc(x, params), lbA1, ubA1,
> optimset('TolX',tol))
>
> Any help would be really appreciated.
>
> Thanks,
> Jude
>
