Hi guys, I am trying to translate a piece of code from Matlab to Julia, please see following first the original Matlab code: options = optimset('Largescale', 'off', 'MaxFunEvals', 500, 'GradObj','on', 'Display','notify'); xp1=fminunc(@stage2_rbf, x, options, X, K, cp, alpha, rho); and now the Jul…

try this code using LinearAlgebra #to use the dot function function myKPCA(X::Matrix{Float64}, #Matrix{Float64} is equivalent to Array{Float64,2} rho_m::Int, #Int is equivalent to Int64 nL_vector::Vector{Int}, #Vector{T} is equivalent to Array{T,1} whidx::UnitRange{Int64}) # myKPCA KPCA…

@longemen3000 so I should try with the procedure you proposed before is it?

Maybe, but the sizes are important, if f is a, 1x1 matrix, the return value should be f[1,1], but if f has more elements, there is a problem in the underlying algorithm. Also, try not returning gradf, as optim is not using it

@longemen3000 I understand what you saying but I am almost sure that ‘f’ is a 1x1 matrix, and also I am almost sure that ‘gradf’ is a 784x1 matrix. I can find for sure the exact sizes in 10or 20 minutes.

Well, in that case, let’s move to give optim a proper function with your gradient. for in this case I recommend something like this : function fg!(F, G, x) f, gradf = stage2_rbf(x) if !(G == nothing) for i=1:length(x) G[i] = gradf[i, 1] end end if !(F == …

@longemen3000 I think you are missing one ‘end’

@longemen3000 I thank you very much for everything Andres, I did not know that I could ask help to optimize my code. I will do that request once I make the code run.

Check this for example [image] In as few lines as possible describe why you love julia New to Julia I was a convert when I rewrote a code that I had originally developed in Matlab. The code I’d written was intended to improve a commercial software I use at work that was p…

@longemen3000 just to confirm for sure the dimensions of ‘f’ are 1x1 and the dimensions of ‘gradf’ are 784x1.

@longemen3000 I did run the code using only your last suggestion (NOT using the suggestion to return f[1, 1]) and I got this error: ERROR: LoadError: UndefVarError: only_fg! not defined Stacktrace: [1] myKPCA(::Array{Float64,2}, ::Int64, ::Array{Int64,1}, ::UnitRange{Int64}) at /home/xxxx_xx/.juli…

Change only_fg! For Optim.only_fg!, or NLSolversBase.only_fg!, maybe is a not exported name

Problem using Optim.optimize

Specific Domains

longemen3000 January 30, 2020, 5:40am 24

Check this for example

Topic		Replies	Views
Performance comparison - Flux.jl's Adam vs Jax's Adam Performance question , package , performance , flux	33	4146	November 3, 2022
Optim.jl vs scipy.optimize once again Performance optim	54	9552	June 10, 2021
Why is this optimization problem so slow in Julia? In R it took 60s Optimization (Mathematical)	31	2166	May 22, 2021
Using optim.jl ( optimization ) - like fminsearch in matlab Optimization (Mathematical) optim	48	6099	November 2, 2017
Maximum Likelihood: Normal Linear Model Optimization (Mathematical)	31	4498	December 25, 2020

Problem using Optim.optimize

Related topics