You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Now that WorkPrecision contains not just a single error and runtimes, but also solver stats like the number of function evaluations, discrete step sizes for non-adaptive solvers, and a range of computed errors, we should provide some convenient functionality to plot these.
Example use-cases I see are
Specifying which error to plot against runtime (currently it plots the one saved in wp.error_estimate, but it could be l2, final, ...)
Plotting error against dts for fixed-step solvers (currently possible with view=:dt_convergence; maybe a more general API would simplify the code?)
Plotting error against stats.nf (as already requested back in 2020 in feval vs error plot #64)
Plotting one error against another one (particularly relevant for the probabilistic solvers in ProbNumDiffEq.jl as to evaluate calibration we plot a calibration metric agains an error)
Plotting tolerances against error
How should the API look?
I'm not quite sure about this, but maybe simple x::Symbol and y::Symbol keyword arguments would be enough?
Something along the lines of
@recipefunctionf(
wp_set::WorkPrecisionSet;
x::Symbol= wp_set.error_estimate,
y::Symbol=:times,
view =:benchmark,
color =nothing
)
and then the x and y data would get filled with the corresponding values.
What do you think, would this make sense? I will try to see if I can get a minimal example working soon., and open a corresponding draft PR.
The text was updated successfully, but these errors were encountered:
Now that WorkPrecision contains not just a single error and runtimes, but also solver stats like the number of function evaluations, discrete step sizes for non-adaptive solvers, and a range of computed errors, we should provide some convenient functionality to plot these.
Example use-cases I see are
wp.error_estimate
, but it could bel2
,final
, ...)dts
for fixed-step solvers (currently possible withview=:dt_convergence
; maybe a more general API would simplify the code?)stats.nf
(as already requested back in 2020 in feval vs error plot #64)How should the API look?
I'm not quite sure about this, but maybe simple
x::Symbol
andy::Symbol
keyword arguments would be enough?Something along the lines of
and then the x and y data would get filled with the corresponding values.
What do you think, would this make sense? I will try to see if I can get a minimal example working soon., and open a corresponding draft PR.
The text was updated successfully, but these errors were encountered: