Something works

ext/EmulatorModelsExt.jl

@@ -12,14 +12,15 @@ import CloudMicrophysics.Parameters as CMP
 function AA.N_activated_per_mode(
     machine::MLJ.Machine,
     ap::CMP.AerosolActivationParameters,
+    ad::CMP.AerosolDistributionType,
     aip::CMP.AirProperties,
     tps::TDP.ThermodynamicsParameters,
     T::FT,
     p::FT,
     w::FT,
     q::TD.PhasePartition{FT},
 ) where {FT <: Real}
-    hygro = mean_hygroscopicity_parameter(ap, ad)
+    hygro = AA.mean_hygroscopicity_parameter(ap, ad)
     return ntuple(Val(AM.n_modes(ad))) do i
         # Model predicts activation of the first mode. So, swap each mode
         # with the first mode repeatedly to predict all activations.

ml_test/Common.jl

@@ -0,0 +1,77 @@
+import CSV
+import DataFrames as DF
+using DataFramesMeta
+
+function get_num_modes(df::DataFrame)
+    i = 1
+    while true
+        if !("mode_$(i)_N" in names(df))
+            return i - 1
+        end
+        i += 1
+    end
+end
+
+function get_num_modes(data_row::NamedTuple)
+    i = 1
+    while true
+        if !(Symbol("mode_$(i)_N") in keys(data_row))
+            return i - 1
+        end
+        i += 1
+    end
+end
+
+function read_aerosol_dataset(
+    dataset_filename::String,
+    Y_name = :mode_1_act_frac_S_interp
+)
+    initial_data = DF.DataFrame(CSV.File(dataset_filename))
+    df = filter(row -> row.S_max > 0 && row.S_max < 0.2, initial_data)
+    selected_columns_X = []
+    num_modes = get_num_modes(df)
+    @info(num_modes)
+    for i in 1:num_modes
+        append!(
+            selected_columns_X,
+            Symbol.([
+                "mode_$(i)_N",
+                "mode_$(i)_mean",
+                "mode_$(i)_stdev",
+                "mode_$(i)_kappa",
+            ]),
+        )
+    end
+    append!(
+        selected_columns_X,
+        [:velocity, :initial_temperature, :initial_pressure],
+    )
+    X = df[:, selected_columns_X]
+    Y = df[:, Y_name]
+    return (X, Y, initial_data)
+end
+
+function preprocess_aerosol_data(X::DataFrame)
+    num_modes = get_num_modes(X)
+    for i in 1:num_modes
+        X = DF.transform(
+            X,
+            Symbol("mode_$(i)_N") => ByRow(log) => Symbol("mode_$(i)_N"),
+        )
+        X = DF.transform(
+            X,
+            Symbol("mode_$(i)_mean") =>
+                ByRow(log) => Symbol("mode_$(i)_mean"),
+        )
+    end
+    X = DF.transform(X, :velocity => ByRow(log) => :velocity)
+    return X
+end
+
+function target_transform(act_frac)
+    return @. atanh(2.0 * 0.99 * (act_frac - 0.5))
+end
+
+function inverse_target_transform(transformed_act_frac)
+    return @. (1.0 / (2.0 * 0.99)) * tanh(transformed_act_frac) + 0.5
+end

ml_test/Project.toml

@@ -1,12 +1,11 @@
 [deps]
-CLIMAParameters = "6eacf6c3-8458-43b9-ae03-caf5306d3d53"
 CSV = "336ed68f-0bac-5ca0-87d4-7b16caf5d00b"
+ClimaParams = "5c42b081-d73a-476f-9059-fd94b934656c"
 CloudMicrophysics = "6a9e3e04-43cd-43ba-94b9-e8782df3c71b"
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
 DataFramesMeta = "1313f7d8-7da2-5740-9ea0-a2ca25f37964"
-EvoTrees = "f6006082-12f8-11e9-0c9c-0d5d367ab1e5"
+Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
-JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819"
 MLJ = "add582a8-e3ab-11e8-2d5e-e98b27df1bc7"
 MLJFlux = "094fc8d1-fd35-5302-93ea-dabda2abf845"
 MLJModels = "d491faf4-2d78-11e9-2867-c94bc002c0b7"