scikitlearn v1.4 support

CondaPkg.toml

@@ -1,4 +1,4 @@
 
 [deps.scikit-learn]
 channel = "conda-forge"
-version = ">=1.2, <1.4"
+version = ">=1.4, <1.5"

src/models/clustering.jl

@@ -30,9 +30,7 @@ meta(AffinityPropagation,
 const AgglomerativeClustering_ = skcl(:AgglomerativeClustering)
 @sk_uns mutable struct AgglomerativeClustering <: MMI.Unsupervised
     n_clusters::Int     = 2::(_ ≥ 1)
-    # replace `affinity` parameter with `metric` when scikit learn releases v1.4
-    affinity::String    = "euclidean"::(_ in ("euclidean", "l1", "l2", "manhattan", "cosine", "precomputed"))
-    #metric::Any = nothing::(_ isa Union{Nothing, Function} || _ in ("euclidean", "l1", "l2", "manhattan", "cosine", "precomputed"))
+    metric::String      = "euclidean"::(_ in ("euclidean", "l1", "l2", "manhattan", "cosine", "precomputed"))
     memory::Any         = nothing
     connectivity::Any   = nothing
     compute_full_tree::Union{String,Bool} = "auto"::(_ isa Bool || _ == "auto")
@@ -187,8 +185,7 @@ const FeatureAgglomeration_ = skcl(:FeatureAgglomeration)
     connectivity::Any      = nothing
     # XXX unclear how to pass a proper callable here; just passing mean = nok
     # pooling_func::Function = mean
-    # replace `affinity` parameter with `metric` when scikit learn releases v1.4
-    affinity::Any          = "euclidean"::(_ isa Function || _ in ("euclidean", "l1", "l2", "manhattan", "cosine",  "precomputed"))
+    metric::Any            = "euclidean"::(_ isa Function || _ in ("euclidean", "l1", "l2", "manhattan", "cosine",  "precomputed"))
     compute_full_tree::Union{String,Bool} = "auto"::(_ isa Bool || _ == "auto")
     linkage::String        = "ward"::(_ in ("ward", "complete", "average", "single"))
     distance_threshold::Option{Float64}   = nothing

src/models/ensemble.jl

@@ -170,6 +170,9 @@ const ExtraTreesRegressor_ = sken(:ExtraTreesRegressor)
     random_state::Any              = nothing
     verbose::Int                   = 0
     warm_start::Bool               = false
+    ccp_alpha::Float64             = 0.0
+    max_samples::Option{Union{Int64,Float64,Nothing}} = nothing
+    monotonic_cst::Option{Union{Vector, Dict}} = nothing
 end
 @sk_feature_importances ExtraTreesRegressor
 MMI.fitted_params(m::ExtraTreesRegressor, (f, _, _)) = (
@@ -216,6 +219,9 @@ const ExtraTreesClassifier_ = sken(:ExtraTreesClassifier)
     verbose::Int                   = 0
     warm_start::Bool               = false
     class_weight::Any              = nothing
+    ccp_alpha::Float64             = 0.0
+    max_samples::Option{Union{Int64,Float64,Nothing}} = nothing
+    monotonic_cst::Option{Union{Vector, Dict}} = nothing
 end
 @sk_feature_importances ExtraTreesClassifier
 MMI.fitted_params(m::ExtraTreesClassifier, (f, _, _)) = (
@@ -367,6 +373,7 @@ const RandomForestRegressor_ = sken(:RandomForestRegressor)
     ccp_alpha::Float64             =0.0::(_ ≥ 0)
     max_samples::Union{Nothing,Float64,Int} =
         nothing::(_ === nothing || (_ ≥ 0 && (_ isa Integer || _ ≤ 1)))
+    monotonic_cst::Option{Union{Vector, Dict}} = nothing
 end
 @sk_feature_importances RandomForestRegressor
 MMI.fitted_params(model::RandomForestRegressor, (f, _, _)) = (
@@ -418,6 +425,7 @@ const RandomForestClassifier_ = sken(:RandomForestClassifier)
     ccp_alpha::Float64             =0.0::(_ ≥ 0)
     max_samples::Union{Nothing,Float64,Int} =
         nothing::(_ === nothing || (_ ≥ 0 && (_ isa Integer || _ ≤ 1)))
+    monotonic_cst::Option{Union{Vector, Dict}} = nothing
 end
 @sk_feature_importances RandomForestClassifier
 MMI.fitted_params(m::RandomForestClassifier, (f, _, _)) = (
@@ -463,7 +471,7 @@ const HistGradientBoostingRegressor_ = sken(:HistGradientBoostingRegressor)
     max_bins::Int                   = 255
     categorical_features::Option{Vector} = nothing
     monotonic_cst::Option{Union{Vector, Dict}} = nothing
-    # interaction_cst
+    interaction_cst::Any            = nothing
     warm_start::Bool                = false
     early_stopping::Union{String, Bool} = "auto"::(_ in ("auto", true, false))
     scoring::String                 = "loss"
@@ -507,7 +515,7 @@ const HistGradientBoostingClassifier_ = sken(:HistGradientBoostingClassifier)
     max_bins::Int                   = 255
     categorical_features::Option{Vector} = nothing
     monotonic_cst::Option{Union{Vector, Dict}} = nothing
-    # interaction_cst
+    interaction_cst::Any            = nothing
     warm_start::Bool                = false
     early_stopping::Union{String, Bool} = "auto"::(_ in ("auto",) || _ isa Bool)
     scoring::String                 = "loss"

src/models/linear-regressors.jl

@@ -124,8 +124,6 @@ const LarsRegressor_ = sklm(:Lars)
 @sk_reg mutable struct LarsRegressor <: MMI.Deterministic
     fit_intercept::Bool      = true
     verbose::Union{Bool,Int} = false
-    # TODO Remove this when python ScikitLearn releases v1.4
-    normalize::Bool          = false
     precompute::Union{Bool,String,AbstractMatrix} = "auto"
     n_nonzero_coefs::Int     = 500::(_ > 0)
     eps::Float64   = eps(Float64)::(_ > 0)
@@ -148,8 +146,6 @@ const LarsCVRegressor_ = sklm(:LarsCV)
     fit_intercept::Bool      = true
     verbose::Union{Bool,Int} = false
     max_iter::Int     = 500::(_ > 0)
-    # TODO Remove this when python ScikitLearn releases v1.4
-    normalize::Bool   = false 
     precompute::Union{Bool,String,AbstractMatrix} = "auto"
     cv::Any           = 5
     max_n_alphas::Int = 1_000::(_ > 0)
@@ -224,8 +220,6 @@ const LassoLarsRegressor_ = sklm(:LassoLars)
     alpha::Float64      = 1.0::(_ ≥ 0) # 0 should be OLS
     fit_intercept::Bool = true
     verbose::Union{Bool, Int} = false
-    # TODO Remove this when python ScikitLearn releases v1.4
-    normalize::Bool     = false
     precompute::Union{Bool,String,AbstractMatrix} = "auto"
     max_iter::Int       = 500::(_ > 0)
     eps::Float64        = eps(Float64)::(_ > 0)
@@ -249,8 +243,6 @@ const LassoLarsCVRegressor_ = sklm(:LassoLarsCV)
     fit_intercept::Bool = true
     verbose::Union{Bool, Int} = false
     max_iter::Int       = 500::(_ > 0)
-    # TODO Remove this when python ScikitLearn releases v1.4
-    normalize::Bool     = false
     precompute::Union{Bool,String,AbstractMatrix} = "auto"
     cv::Any             = 5
     max_n_alphas::Int   = 1_000::(_ > 0)
@@ -278,8 +270,6 @@ const LassoLarsICRegressor_ = sklm(:LassoLarsIC)
     criterion::String   = "aic"::(_ in ("aic","bic"))
     fit_intercept::Bool = true
     verbose::Union{Bool, Int} = false
-    # TODO Remove this when python ScikitLearn releases v1.4
-    normalize::Bool     = false
     precompute::Union{Bool,String,AbstractMatrix} = "auto"
     max_iter::Int       = 500::(_ > 0)
     eps::Float64        = eps(Float64)::(_ > 0.0)
@@ -315,7 +305,6 @@ const OrthogonalMatchingPursuitRegressor_ = sklm(:OrthogonalMatchingPursuit)
     n_nonzero_coefs::Option{Int} = nothing
     tol::Option{Float64} = nothing
     fit_intercept::Bool  = true
-    normalize::Bool      = false
     precompute::Union{Bool,String,AbstractMatrix} = "auto"
 end
 MMI.fitted_params(model::OrthogonalMatchingPursuitRegressor, (fitresult, _, _)) = (
@@ -329,8 +318,6 @@ const OrthogonalMatchingPursuitCVRegressor_ = sklm(:OrthogonalMatchingPursuitCV)
 @sk_reg mutable struct OrthogonalMatchingPursuitCVRegressor <: MMI.Deterministic
     copy::Bool            = true
     fit_intercept::Bool   = true
-    # TODO Remove this when python ScikitLearn releases v1.4
-    normalize::Bool       = false
     max_iter::Option{Int} = nothing::(_ === nothing||_ > 0)
     cv::Any               = 5
     n_jobs::Option{Int}   = 1