Skip to content

gragusa/CsminWel.jl

Repository files navigation

CsminWel.jl

Build Status Coverage Status codecov.io

Interface to Chris Sims' csminwel optimization code. The code borrows from DSGE.jl, but it is adapted to be compatible with Optim.jl's API. When the derivative of the minimand is not supplied, either Finite Difference of Forward Automatic Differentiation derivatives are used.

Differently from the solvers in Optim.jl, Csminwel returns an estimate of the inverse of the Hessian at the solution.

#=
Maximizing loglikelihood logistic models
=#
using CsminWel, StatsFuns, Random
Random.seed!(1)
x = [ones(200) randn(200,4)]
y = [rand() < 0.5 ? 1. : 0. for j in 1:200]

function loglik(beta)
    xb = x*beta
    sum(-y.*xb + log1pexp.(xb))
end

function dloglik(beta)
    xb = x*beta
    px = logistic.(xb)
    -x'*(y.-px)
end

function g!(stor, beta)
    copyto!(stor, dloglik(beta))
end

res1 = optimize(loglik, g!, zeros(5), BFGS())
res2 = optimize(loglik, g!, zeros(5), Csminwel())

## With finite-difference derivative
res3 = optimize(loglik, zeros(5), Csminwel())

## With forward AD derivative
res4 = optimize(Optim.OnceDifferentiable(loglik, zeros(5), autodiff = :forward), zeros(5), Csminwel())

## inverse Hessian
res2.invH

About

Optimization in Julia

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages