For a 1.5 release (#154)

* add `feature_importances` stub  (#148)

* add intrinsic_importances stub and set fallback to
othing.

* fix error in intrinsic_importances docstring.

* rename intrinsic_importances method to eature_importances.

* remove fallback for eature_importances.

* Update src/model_api.jl

Co-authored-by: Anthony Blaom, PhD <[email protected]>

* bump 1.5

Co-authored-by: Okon Samuel <[email protected]>

Author: ablaom

Project.toml

@@ -1,7 +1,7 @@
 name = "MLJModelInterface"
 uuid = "e80e1ace-859a-464e-9ed9-23947d8ae3ea"
 authors = ["Thibaut Lienart and Anthony Blaom"]
-version = "1.4.4"
+version = "1.5"
 
 [deps]
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"

src/MLJModelInterface.jl

@@ -30,6 +30,7 @@ const MODEL_TRAITS = [
     :hyperparameter_ranges,
     :iteration_parameter,
     :supports_training_losses,
+    :reports_feature_importances,
     :deep_properties
 ]
 

src/model_api.jl

@@ -40,7 +40,7 @@ implement this method to return an `AbstractVector` of the losses
 in historical order. If the model calculates scores instead, then the
 sign of the scores should be reversed.
 
-The following trait overload is alse required:
+The following trait overload is also required:
 `supports_training_losses(::Type{<:M}) = true`
 
 """
@@ -164,3 +164,20 @@ function restore end
 some meta-models may choose to implement the `evaluate` operations
 """
 function evaluate end
+
+"""
+    feature_importances(model::M, fitresult, report)
+
+For a given `model` of model type `M` supporting intrinsic feature importances, calculate 
+the feature importances from the model's `fitresult` and `report` as an 
+abstract vector of `feature::Symbol => importance::Real` pairs 
+(e.g `[:gender =>0.23, :height =>0.7, :weight => 0.1]`).  
+
+The following trait overload is also required:
+`reports_feature_importances(::Type{<:M}) = true`
+
+If for some reason a model is sometimes unable to report feature importances then
+`feature_importances` should return all importances as 0.0, as in 
+`[:gender =>0.0, :height =>0.0, :weight => 0.0]`. 
+"""
+function feature_importances end

test/model_api.jl

@@ -5,6 +5,7 @@ end
     f0::Int
 end
 
+
 mutable struct APIx1 <: Static end
 
 @testset "selectrows(model, data...)" begin
@@ -30,12 +31,18 @@ end
     s1 = APIx1()
     @test fit(s1, 1, 0) == (nothing, nothing, nothing)
 
-    #update fallback = fit
+    # update fallback = fit
     @test update(m0, 1, 5, nothing, randn(2), 5) == (5, nothing, nothing)
 
     # training losses:
     f, c, r = MLJModelInterface.fit(m0, 1, rand(2), rand(2))
     @test M.training_losses(m0, r) === nothing
+    
+    # intrinsic_importances
+    f, c, r = MLJModelInterface.fit(m0, 1, rand(2), rand(2))
+    MLJModelInterface.reports_feature_importances(::Type{APIx0}) = true
+    MLJModelInterface.feature_importances(::APIx0, fitresult, report) = [:a=>0, :b=>0]
+    @test MLJModelInterface.feature_importances(m0, f, r) == [:a=>0, :b=>0]
 end
 
 struct DummyUnivariateFinite end