Aerosol activation emulators: GP, ET, and ARG-informed #349
Conversation
…d training with informed data from ARG
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #349 +/- ##
==========================================
+ Coverage 98.11% 98.57% +0.46%
==========================================
Files 35 35
Lines 1165 1195 +30
==========================================
+ Hits 1143 1178 +35
+ Misses 22 17 -5 ☔ View full report in Codecov by Sentry. |
ext/Common.jl
Outdated
| @@ -68,6 +71,67 @@ function preprocess_aerosol_data(X::DataFrame) | |||
| return X | |||
| end | |||
|
|
|||
| function get_ARG_act_frac(data_row::NamedTuple) | |||
|
|
|||
| FT = Float32 | |||
There was a problem hiding this comment.
Would it be possible to pass in the FT and the parameter structs as arguments, instead of defining them here?
| pv0 = TD.saturation_vapor_pressure(tps, FT(T), TD.Liquid()) | ||
| vapor_mix_ratio = pv0 / TD.Parameters.molmass_ratio(tps) / (p - pv0) | ||
| q_vap = vapor_mix_ratio / (vapor_mix_ratio + 1) | ||
| q = TD.PhasePartition(FT(q_vap), FT(0), FT(0)) |
There was a problem hiding this comment.
Are we sure there is never any cloud condensate in this data?
test/aerosol_activation_emulators.jl
Outdated
| FT = Float32 | ||
|
|
||
| # If the ML model already exists load it in. | ||
| # If it does not exist, train a NN |
There was a problem hiding this comment.
| # If it does not exist, train a NN | |
| # If it does not exist, train |
| ml_model = "NN", | ||
| machine_name = "2modal_nn_machine_naive.jls", | ||
| ) | ||
|
|
There was a problem hiding this comment.
Could we encapsulate individual tests with
TT.@testset "0M_microphysics" begin
end
blocks? I think this way they display nicer in the test summary in the CI
There was a problem hiding this comment.
Thank you for doing this! My two asks would be to:
- not hardcode the
FTand parameter structs inCommon.jl - add some small documentation page about the available extensions. Apologies I didn't do it with the initial NN extension. I was planning to add it when we have the other guys. I was thinking to add a short section in the Aerosol Activation part of our docs and mention there that we have those additional models. (We can add plots, once we have them for the paper)
|
Actually, one more thought. Is there a way to add the |
|
I tried loading the extension in docs to make Julia understand we have this extra function, but it doesn't add it to API. Maybe documenter goes over the specific files of each module to get the documentations and does not do it with the extensions. |
…mentation for the new ML extensions
This PR follows #318 with adding the remaining parts of the work of @mikhailmints. Here we add Gaussian processes and Evotrees. Also, informing data with ARG for training is enabled.