Benchmarking with different ODE solvers
Initial benchmarking code for the toy example using different deSolve::ode
methods can be found in the benchmarks
branch in the "tests/benchmarks"
folder.
The result for this case study was that the fastet method, when scaling reaches and habitats (number of ODE vars), was "adams"
method by a small margin of ca. 20% over multiple other solvers (cf. benchmarks.toy.reach_and_habitat.zip), including "lsoda"
method, which is currently a default method used in run.streambugs
call.
Try:
-
same benchmarks with the extended example, prepared within !1 (merged) . -
other than deSolve pkgs, e.g. sundials via cvode
R pkg, or ode23s via pracma R pkg, or dedicated CRN solver (e.g. odeSD C version); tuning: tolerance; consideration: support for discrete perturbations/inputs/events.
Est: 2d
Edited by Mikołaj Rybiński