Using NLopt for maximum likelihood estimation

New to Julia
1

Hi all,
I am an R user now beginning with Julia. As a start I’m trying to figure out how I can optimize a likelihood function in Julia. However, i cannot get my first try to work. Could anyone help me figuring out what it is that I do wrong?

Thank you!

using NLopt
using Distributions

y =  [1, 2, 1, 2, 3]

function loglnorm(par)
    ll = 0.0
    for i = 1:length(y)
        ll -= logpdf(Normal(par[1], par[2]), y[i])
    end
    return ll
end

opt = Opt(:LN_BOBYQA, 2)
opt.lower_bounds = [-Inf, 0.0]
 opt.xtol_rel = 1e-4
opt.min_objective = loglnorm
(minf,minx,ret) = optimize!(opt, [2.0, 1.0])
numevals = opt.numevals
println("obj: $minf, par: $minx, evals: $numevals, return: $ret)")

The return value from optimize always return: “FORCED_STOP”.

2

NLopt always expects two arguments objective function and if not given it silently fails because exception at the NLopt library calls are not caught but causes forced return. I also had the same problem. Make loglnorm

function loglnorm(par, dummygrad)
    ll = 0.0
    for i = 1:length(y)
        ll -= logpdf(Normal(par[1], par[2]), y[i])
    end
    return ll
end
3

Thanks a lot! That worked.