ZerothOrderOptimizer.jl

module A

    using Optim
    import Optim: ZerothOrderOptimizer, ZerothOrderState, initial_state, update_state!, trace!, assess_convergence, AbstractOptimizerState, update!, value

    struct RandomSampler <: ZerothOrderOptimizer
        σ::Float64
    end

    RandomSampler(; σ=1e-1) = RandomSampler(σ)

    type RandomSamplerState{T,N} <: ZerothOrderState
        x::Array{T,N}
        x_previous::Array{T,N}
        f_x_previous::T
    end

    function initial_state(method::RandomSampler, options, d, initial_x)
        # prepare cache variables etc here
        RandomSamplerState(initial_x,initial_x,Inf)
    end

    function update_state!{T}(d, state::RandomSamplerState{T}, method::RandomSampler)
        
        x = state.x + method.σ*randn(length(state.x))
        f_x = value(d, x)
        
        if f_x < state.f_x_previous
            state.f_x_previous = f_x
            state.x = x
        end
        
        false # should the procedure force quit?
    end
    
    function trace!(tr, d, state, iteration, method::RandomSampler, options)
            dt = Dict()
            if options.extended_trace
                dt["x"] = copy(state.x)
            end
            g_norm = 0.0
            update!(tr,
            iteration,
            value(d),
            g_norm,
            dt,
            options.store_trace,
            options.show_trace,
            options.show_every,
            options.callback)
    end
    
    function assess_convergence(state::RandomSamplerState, d, options)
        Optim.default_convergence_assessment(state, d, options)
    end
    
end


f = x -> sum(x.^2) + π
mfit = optimize(f,rand(2),A.RandomSampler(),Optim.Options(iterations=500,store_trace=false,extended_trace=false))