and now, while sol1, sol2, and sol3 are different, they will yield reproducibly the same result each time I rerun the above snippet (useful to e.g. remove any randomness from tests). Note, this si different from setting a seed, where if I do:
StochasticDiffEq does not have a feature like this (and is only able to set seeds, however, it would be useful.
It is possible to partially circumvent this by setting different seeds for each sol (potentially depending on the StableRNG). However, long-term, being able to provide a stable rng to a SDEProblem directly would be ideal.
I think the bigger issue here is providing a simple way for users to set the rng used internally, i.e. via the problem or solve call, which would also fix this issue.
For JumpProcesses's
JumpProblem
s you can doand now, while
sol1
,sol2
, andsol3
are different, they will yield reproducibly the same result each time I rerun the above snippet (useful to e.g. remove any randomness from tests). Note, this si different from setting aseed
, where if I do:sol1
,sol2
, andsol3
will be identical.StochasticDiffEq does not have a feature like this (and is only able to set
seed
s, however, it would be useful.It is possible to partially circumvent this by setting different
seed
s for each sol (potentially depending on the StableRNG). However, long-term, being able to provide a stable rng to aSDEProblem
directly would be ideal.