Merge pull request #149 from JuliaAI/dev

For a 0.6.12 release

Project.toml

@@ -1,7 +1,7 @@
 name = "MLJTuning"
 uuid = "03970b2e-30c4-11ea-3135-d1576263f10f"
 authors = ["Anthony D. Blaom <anthony.blaom@gmail.com>"]
-version = "0.6.11"
+version = "0.6.12"
 
 [deps]
 ComputationalResources = "ed09eef8-17a6-5b46-8889-db040fac31e3"
@@ -17,7 +17,7 @@ RecipesBase = "3cdcf5f2-1ef4-517c-9805-6587b60abb01"
 ComputationalResources = "0.3"
 Distributions = "0.22,0.23,0.24, 0.25"
 LatinHypercubeSampling = "1.7.2"
-MLJBase = "0.18.15"
+MLJBase = "0.18.19"
 MLJModelInterface = "0.4.1, 1.1.1"
 ProgressMeter = "1.7.1"
 RecipesBase = "0.8,0.9,1"

src/strategies/latin_hypercube.jl

@@ -1,12 +1,12 @@
 """
-LatinHypercube(gens = 1,
-               popsize = 100,
-               ntour = 2,
-               ptour = 0.8.,
-               interSampleWeight = 1.0,
-               ae_power = 2,
-               periodic_ae = false,
-               rng=Random.GLOBAL_RNG)
+    LatinHypercube(gens = 1,
+                   popsize = 100,
+                   ntour = 2,
+                   ptour = 0.8.,
+                   interSampleWeight = 1.0,
+                   ae_power = 2,
+                   periodic_ae = false,
+                   rng=Random.GLOBAL_RNG)
 
 Instantiate grid-based hyperparameter tuning strategy using the
 library [LatinHypercubeSampling.jl](https://github.com/MrUrq/LatinHypercubeSampling.jl).

src/tuned_models.jl

@@ -69,7 +69,7 @@ const EitherTunedModel{T,M} =
                              range=nothing,
                              measure=nothing,
                              n=default_n(tuning, range),
-                             operation=predict,
+                             operation=nothing,
                              other_options...)
 
 Construct a model wrapper for hyper-parameter optimization of a
@@ -80,7 +80,7 @@ hyper-parameters are to be mutated.
                              resampling=Holdout(),
                              measure=nothing,
                              n=default_n(tuning, range),
-                             operation=predict,
+                             operation=nothing,
                              other_options...)
 
 Construct a wrapper for multiple `models`, for selection of an optimal
@@ -185,10 +185,10 @@ plus other key/value pairs specific to the `tuning` strategy.
 - `repeats=1`: for generating train/test sets multiple times in
   resampling; see [`evaluate!`](@ref) for details
 
-- `operation=predict`: operation to be applied to each fitted model;
-  usually `predict` but `predict_mean`, `predict_median` or
-  `predict_mode` can be used for `Probabilistic` models, if
-  the specified measures are `Deterministic`
+- `operation`/`operations` - One of
+  $(MLJBase.PREDICT_OPERATIONS_STRING), or a vector of these of the
+  same length as `measure`/`measures`. Automatically inferred if left
+  unspecified.
 
 - `range`: range object; tuning strategy documentation describes
   supported types
@@ -227,7 +227,8 @@ function TunedModel(; model=nothing,
                     measures=nothing,
                     measure=measures,
                     weights=nothing,
-                    operation=predict,
+                    operations=nothing,
+                    operation=operations,
                     ranges=nothing,
                     range=ranges,
                     selection_heuristic=NaiveSelection(),

test/tuned_models.jl

@@ -294,4 +294,20 @@ end
     @test all(==(per_folds[1]), per_folds)
 end
 
+@testset "deterministic metrics for probabilistic models" begin
+
+    # https://github.com/JuliaAI/MLJBase.jl/pull/599 allows mix of
+    # deterministic and probabilistic metrics:
+    X, y = MLJBase.make_blobs()
+    model = DecisionTreeClassifier()
+    range = MLJBase.range(model, :max_depth, values=[1,2])
+    tmodel = TunedModel(model=model,
+                        range=range,
+                        measures=[MisclassificationRate(),
+                                  LogLoss()])
+    mach = machine(tmodel, X, y)
+    @test_logs fit!(mach, verbosity=0)
+
+end
+
 true