DiffEqNoiseProcess.jl
DiffEqNoiseProcess.jl copied to clipboard
correlated Wiener tests do not pass on windows
Correlated Wiener Process: Test Failed at C:\Users\arno\Desktop\SciML\DiffEqNoiseProcess.jl\test\correlated.jl:36
Expression: ≈(Γ, cov(sol, dims = 2) / dt, rtol = 0.01)
Evaluated: [1.0 0.3; 0.3 1.0] ≈ [1.4071833818984558 0.4647373260398336; 0.4647373260398336 1.4034111425410771] (rtol=0.01)
Stacktrace:
[1] macro expansion
@ C:\Users\arno\.julia\juliaup\julia-1.9.4+0.x64.w64.mingw32\share\julia\stdlib\v1.9\Test\src\Test.jl:478 [inlined]
[2] macro expansion
@ C:\Users\arno\Desktop\SciML\DiffEqNoiseProcess.jl\test\correlated.jl:36 [inlined]
[3] macro expansion
@ C:\Users\arno\.julia\juliaup\julia-1.9.4+0.x64.w64.mingw32\share\julia\stdlib\v1.9\Test\src\Test.jl:1498 [inlined]
[4] top-level scope
@ C:\Users\arno\Desktop\SciML\DiffEqNoiseProcess.jl\test\correlated.jl:3
Is the tolerance just slightly off? We can run for more iterations to make the value converge better.
1 vs 1.4 ; 0.3 vs 0.4 I would not call slightly. You can see it in the output.
What about with 10_000_000 trajectories?
28.376316 seconds (980.54 M allocations: 71.932 GiB, 22.86% gc time, 14.08% compilation time)
Correlated Wiener Process: Test Failed at C:\Users\arno\Desktop\SciML\DiffEqNoiseProcess.jl\test\correlated.jl:36
Expression: ≈(Γ, cov(sol, dims = 2) / dt, rtol = 0.01)
Evaluated: [1.0 0.3; 0.3 1.0] ≈ [1.770707414607028 0.6011729477076763; 0.6011729477076763 1.7710078697122533] (rtol=0.01)
oh that's weird. Just on windows?
Yes, not on linux, did not test mac.