I tested the example in https://docs.juliadiffeq.org/stable/analysis/global_sensitivity/#GSA-examples-1 . As expected it all went well. This example, calculates sensitivity with regard to the mean over time on u1 and the maximum over time on u2. I am interesting on looking how sensitivities change…

[image] alewolf: If I change the function to include sensitivities only fro u1 or u2 (below) it works, It wants a vector. It’s a vector → vector function which then computes the global sensitivity of each output w.r.t.the inputs. You could use that to reshape to a timeseries on your own thoug…

On the same page- I tried to run the sobol parallel code you have in the documentation. it did not work. There mightbe a bug there. N = 10000 lb = [1.0, 1.0, 1.0, 1.0] ub = [5.0, 5.0, 5.0, 5.0] sampler = SobolSample() A,B = QuasiMonteCarlo.generate_design_matrices(N,lb,ub,sampler) function f(du,u…

f2 = function (p) prob_func(prob,i,repeat) = remake(prob;p=p[i,:]) ensemble_prob = EnsembleProblem(prob,prob_func=prob_func) sol = solve(ensemble_prob,Tsit5(),EnsembleThreads();saveat=t) # Case1 # sol = solve(ensemble_prob,Tsit5(),EnsembleThreads();saveat=t, trajectories=500 ) # Case2 # …

oh that should be sol = solve(ensemble_prob,Tsit5(),EnsembleThreads();saveat=t,trajectories=size(p,1)).

[image] ChrisRackauckas: solve(ensemble_prob,Tsit5(),EnsembleThreads();saveat=t,trajectories=size(p,1)) Sorry to keep at it I tried the fix you suggested - Still does not work for me, I am showing full code and error below. Still get bounds error: function f(du,u,p,t) du[1] = p[1]*u[1] -…

Should the code have a double loop? This does not crash, but I am not 100% sure I am looking at the right thing anymore. Results are veru different from serial analysis function f(du,u,p,t) # @inbounds begin du[1] = p[1]*u[1] - p[2]*u[1]*u[2] #prey du[2] = -p[3]*u[2] + p[4]*u[1]*u[2] #predat…

oh I think I had the indexing flipped: using OrdinaryDiffEq, DiffEqSensitivity function f(du,u,p,t) du[1] = p[1]*u[1] - p[2]*u[1]*u[2] #prey du[2] = -p[3]*u[2] + p[4]*u[1]*u[2] #predator end u0 = [1.0;1.0] tspan = (0.0,10.0) p = [1.5,1.0,3.0,1.0] prob = ODEProblem(f,u0,tspan,p) t = collect(rang…

Thanks! I’ll try this as soon as I have 5 min today. I still get quite confused with indexing for parallel or ensemble computation. So I appreciate you taking the time to help :slight_smile: Works perfect! A

Last question on this subject - I am able to tun Sobol on my real model (44 equations, 65 params) As long as I keep N below 10,000. This seems like a small number for such a big set of params. When I tried 50000, my RAM collapsed (usage 100%). I am using a system, provided at work, with 64GB RAM. …

Global sensitivity analysis - following time series

Specific Domains Modelling & Simulations

ChrisRackauckas March 12, 2020, 12:15pm 4

If you batch you have to change your function.

Topic		Replies	Views
Parallelize GSA on continuous outputs Modelling & Simulations diffeq	12	490	March 16, 2023
Global sensitivity analysis in a large parameter space - computational cost Modelling & Simulations	15	1448	November 26, 2024
How to plot the parallelized example in "Parallelized Morris and Sobol Sensitivity Analysis of an ODE" Modelling & Simulations globalsensitivityjl	1	68	January 13, 2026
Morris Vs Sobol Method Global Sensitivity analysis Statistics	25	2939	July 7, 2022
Global sensitivity analysis - Morris Analysis New to Julia	21	2254	October 12, 2020

Global sensitivity analysis - following time series

Related topics