I am solving a system of equations with the iterative proportional fitting procedure . That means that at each iteration I solve for a subset of the variables μ0m, then hold those fixed and solve for the other subset of variables μ0f. I am storing the residuals of the system in preallocated arrays re…

I believe you accidentally swapped the first and second argument of jacobian!() .

With this algorithm function ipfp_ad(init, sys!; Sm=Sm, Sf=Sf, Π=Π, method=:trust_region, outmaxiter=5_000, outtol=1e-8, opts...) μ0f0 = vec(init) μ0m1 = similar(init) μ0f1 = similar(init) d = Inf i = 0 while (i <= outmaxiter) && (outtol < d) lfunm! = (resm, μ0m) -> …

sys_jacf! is not a function, it is an array. However, nlsolve expects a function to calculate the jacobian. Please also provide the full stack trace. From the error it impossible to figure out on which line of your code the problem is

Here I post the full script: using NLsolve using LinearAlgebra using ForwardDiff const szA = 9 const szE = 2 const szG = 2 function sys!(resm, resf, μ0m, μ0f; Sm=Sm, Sf=Sf, Π=Π) μ0 = reshape(vcat(vec(μ0m), vec(μ0f)), (szA, szE, szG)) μ0m′ = μ0[:, :, 1] μ0f′ = μ0[:, :, 2] μ0m′[1, …

[image] josuagrw: sys_jacf! is not a function, it is an array. so this should do it? sys_jacm! = (resm, x) -> ForwardDiff.jacobian!(resm, lfunm!, x)

I changed the lines defining the Jacobian: using NLsolve using LinearAlgebra using ForwardDiff const szA = 9 const szE = 2 const szG = 2 const β = 0.92 const δ = 0.10 function sys!(resm, resf, μ0m, μ0f; Sm=Sm, Sf=Sf, Π=Π) μ0 = reshape(vcat(vec(μ0m), vec(μ0f)), (szA, szE, szG)) μ0m′ = μ0[:…

You’re still calling jacobian!() incorrectly. For your case you need to provide 4 arguments. Please read the documentation—it is well explained there.

How to implement AD for this function?

General Usage

amrods May 7, 2021, 9:03am 10

yeah, i missed that bang at the end, see A bug in ForwardDiff? - #4 by amrods

Topic		Replies	Views
How to autodifferentiate the results of NLsolve? General Usage autodiff	24	2659	March 8, 2023
Obtain jacobian used by NLsolve General Usage question , nlsolve	2	355	July 12, 2023
Expressions for partial derivatives using AD tools Optimization (Mathematical)	2	535	August 29, 2020
AD step for iterative solution Numerics question	11	1864	July 17, 2019
NonlinearSolve uses AD Jacobian even when I specify it analytically Numerics nonlinearsolve	3	207	November 14, 2024

How to implement AD for this function?

Related topics