MLJBase->MLJModelInterface

Author: tlienart

Project.toml

@@ -8,20 +8,20 @@ DocStringExtensions = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
 IterativeSolvers = "42fd0dbc-a981-5370-80f2-aaf504508153"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 LinearMaps = "7a12625a-238d-50fd-b39a-03d52299707e"
-MLJBase = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
+MLJModelInterface = "e80e1ace-859a-464e-9ed9-23947d8ae3ea"
 Optim = "429524aa-4258-5aef-a3af-852621145aeb"
 Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 
 [compat]
 DocStringExtensions = "^0.8"
 IterativeSolvers = "^0.8"
-LinearMaps = "^2.5"
-MLJBase = "^0.9"
-Optim = "^0.19"
+LinearMaps = "^2.6"
+MLJModelInterface = "^0.1"
+Optim = "^0.20"
 Parameters = "^0.12"
 Tables = "^0.2"
-julia = "^1.0.0"
+julia = "^1"
 
 [extras]
 DelimitedFiles = "8bb1440f-4735-579b-a4ab-409b98df4dab"

src/MLJLinearModels.jl

@@ -6,12 +6,12 @@ import LinearMaps: LinearMap
 import IterativeSolvers: cg
 import Optim
 
-import MLJBase
+import MLJModelInterface
 
 import Base.+, Base.-, Base.*, Base./, Base.convert
 
-const AVR = AbstractVector{<:Real}
-
+const MMI       = MLJModelInterface
+const AVR       = AbstractVector{<:Real}
 const Option{T} = Union{Nothing,T}
 
 include("scratchspace.jl")

src/mlj/classifiers.jl

@@ -2,7 +2,7 @@
    LOGISTIC CLASSIFIER
    =================== =#
 
-@with_kw_noshow mutable struct LogisticClassifier <: MLJBase.Probabilistic
+@with_kw_noshow mutable struct LogisticClassifier <: MMI.Probabilistic
     lambda::Real             = 1.0
     gamma::Real              = 0.0
     penalty::SymStr          = :l2
@@ -23,7 +23,7 @@ descr(::Type{LogisticClassifier}) = "Classifier corresponding to the loss functi
    MULTINOMIAL CLASSIFIER
    ====================== =#
 
-@with_kw_noshow mutable struct MultinomialClassifier <: MLJBase.Probabilistic
+@with_kw_noshow mutable struct MultinomialClassifier <: MMI.Probabilistic
     lambda::Real             = 1.0
     gamma::Real              = 0.0
     penalty::SymStr          = :l2

src/mlj/interface.jl

@@ -7,17 +7,18 @@ const SymStr = Union{Symbol,String}
 include("regressors.jl")
 include("classifiers.jl")
 
-const REG_MODELS = (LinearRegressor, RidgeRegressor, LassoRegressor, ElasticNetRegressor,
-                    RobustRegressor, HuberRegressor, QuantileRegressor, LADRegressor)
+const REG_MODELS = (LinearRegressor, RidgeRegressor, LassoRegressor,
+                    ElasticNetRegressor, RobustRegressor, HuberRegressor,
+                    QuantileRegressor, LADRegressor)
 const CLF_MODELS = (LogisticClassifier, MultinomialClassifier)
 const ALL_MODELS = (REG_MODELS..., CLF_MODELS...)
 
 #= ==========
    REGRESSORS
    ========== =#
 
-function MLJBase.fit(m::Union{REG_MODELS...}, verb::Int, X, y)
-    Xmatrix = MLJBase.matrix(X)
+function MMI.fit(m::Union{REG_MODELS...}, verb::Int, X, y)
+    Xmatrix = MMI.matrix(X)
     reg     = glr(m)
     solver  = m.solver === nothing ? _solver(reg, size(Xmatrix)) : m.solver
     # get the parameters
@@ -26,9 +27,9 @@ function MLJBase.fit(m::Union{REG_MODELS...}, verb::Int, X, y)
     return θ, nothing, NamedTuple{}()
 end
 
-MLJBase.predict(m::Union{REG_MODELS...}, θ, Xnew) = apply_X(MLJBase.matrix(Xnew), θ)
+MMI.predict(m::Union{REG_MODELS...}, θ, Xnew) = apply_X(MMI.matrix(Xnew), θ)
 
-function MLJBase.fitted_params(m::Union{REG_MODELS...}, θ)
+function MMI.fitted_params(m::Union{REG_MODELS...}, θ)
     m.fit_intercept && return (coefs = θ[1:end-1], intercept = θ[end])
     return (coefs = θ, intercept = nothing)
 end
@@ -37,10 +38,10 @@ end
    CLASSIFIERS
    =========== =#
 
-function MLJBase.fit(m::Union{CLF_MODELS...}, verb::Int, X, y)
-    Xmatrix  = MLJBase.matrix(X)
-    yplain   = convert.(Int, MLJBase.int(y))
-    classes  = MLJBase.classes(y[1])
+function MMI.fit(m::Union{CLF_MODELS...}, verb::Int, X, y)
+    Xmatrix  = MMI.matrix(X)
+    yplain   = convert.(Int, MMI.int(y))
+    classes  = MMI.classes(y[1])
     nclasses = length(classes)
     if nclasses == 2
         # recode
@@ -58,21 +59,21 @@ function MLJBase.fit(m::Union{CLF_MODELS...}, verb::Int, X, y)
     return (θ, c, classes), nothing, NamedTuple{}()
 end
 
-function MLJBase.predict(m::Union{CLF_MODELS...}, (θ, c, classes), Xnew)
-    Xmatrix = MLJBase.matrix(Xnew)
+function MMI.predict(m::Union{CLF_MODELS...}, (θ, c, classes), Xnew)
+    Xmatrix = MMI.matrix(Xnew)
     preds   = apply_X(Xmatrix, θ, c)
     # binary classification
     if c == 1
         preds  .= sigmoid.(preds)
         preds   = hcat(1.0 .- preds, preds) # scores for -1 and 1
-        return [MLJBase.UnivariateFinite(classes, preds[i, :]) for i in 1:size(Xmatrix,1)]
+        return [MMI.UnivariateFinite(classes, preds[i, :]) for i in 1:size(Xmatrix,1)]
     end
     # multiclass
     preds .= softmax(preds)
-    return [MLJBase.UnivariateFinite(classes, preds[i, :]) for i in 1:size(Xmatrix,1)]
+    return [MMI.UnivariateFinite(classes, preds[i, :]) for i in 1:size(Xmatrix,1)]
 end
 
-function MLJBase.fitted_params(m::Union{CLF_MODELS...}, (θ, c, classes))
+function MMI.fitted_params(m::Union{CLF_MODELS...}, (θ, c, classes))
     if c > 1
         if m.fit_intercept
             W = reshape(θ, div(length(θ), c), c)
@@ -90,7 +91,7 @@ end
    METADATA FOR ALL MODELS
    ======================= =#
 
-MLJBase.metadata_pkg.(ALL_MODELS,
+MMI.metadata_pkg.(ALL_MODELS,
     name="MLJLinearModels",
     uuid="6ee0df7b-362f-4a72-a706-9e79364fb692",
     url="https://github.com/alan-turing-institute/MLJLinearModels.jl",
@@ -100,21 +101,21 @@ MLJBase.metadata_pkg.(ALL_MODELS,
 
 descr_(M) = descr(M) *
     "\n→ based on [MLJLinearModels](https://github.com/alan-turing-institute/MLJLinearModels.jl)" *
-    "\n→ do `@load $(MLJBase.name(M)) pkg=\"MLJLinearModels\" to use the model.`" *
-    "\n→ do `?$(MLJBase.name(M))` for documentation."
-lp_(M) = "MLJLinearModels.$(MLJBase.name(M))"
+    "\n→ do `@load $(MMI.name(M)) pkg=\"MLJLinearModels\" to use the model.`" *
+    "\n→ do `?$(MMI.name(M))` for documentation."
+lp_(M) = "MLJLinearModels.$(MMI.name(M))"
 
 for M in REG_MODELS
-    MLJBase.metadata_model(M,
-        input=MLJBase.Table(MLJBase.Continuous),
-        target=AbstractVector{MLJBase.Continuous},
+    MMI.metadata_model(M,
+        input=MMI.Table(MMI.Continuous),
+        target=AbstractVector{MMI.Continuous},
         weights=false,
         descr=descr_(M), path=lp_(M))
 end
 for M in CLF_MODELS
-    MLJBase.metadata_model(M,
-        input=MLJBase.Table(MLJBase.Continuous),
-        target=AbstractVector{<:MLJBase.Finite},
+    MMI.metadata_model(M,
+        input=MMI.Table(MMI.Continuous),
+        target=AbstractVector{<:MMI.Finite},
         weights=false,
         descr=descr_(M), path=lp_(M))
 end

src/mlj/regressors.jl

@@ -2,7 +2,7 @@
    LINEAR REGRESSOR (OLS)
    ====================== =#
 
-@with_kw_noshow mutable struct LinearRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct LinearRegressor <: MMI.Deterministic
     fit_intercept::Bool    = true
     solver::Option{Solver} = nothing
 end
@@ -15,7 +15,7 @@ descr(::Type{LinearRegressor}) = "Regression with objective function ``|Xθ - y|
    RIDGE REGRESSOR
    =============== =#
 
-@with_kw_noshow mutable struct RidgeRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct RidgeRegressor <: MMI.Deterministic
     lambda::Real             = 1.0
     fit_intercept::Bool      = true
     penalize_intercept::Bool = false
@@ -31,7 +31,7 @@ descr(::Type{RidgeRegressor}) = "Regression with objective function ``|Xθ - y|
    LASSO REGRESSOR
    =============== =#
 
-@with_kw_noshow mutable struct LassoRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct LassoRegressor <: MMI.Deterministic
     lambda::Real             = 1.0
     fit_intercept::Bool      = true
     penalize_intercept::Bool = false
@@ -47,7 +47,7 @@ descr(::Type{LassoRegressor}) = "Regression with objective function ``|Xθ - y|
    ELASTIC NET REGRESSOR
    ===================== =#
 
-@with_kw_noshow mutable struct ElasticNetRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct ElasticNetRegressor <: MMI.Deterministic
     lambda::Real             = 1.0
     gamma::Real              = 0.0
     fit_intercept::Bool      = true
@@ -65,7 +65,7 @@ descr(::Type{ElasticNetRegressor}) = "Regression with objective function ``|Xθ
    ROBUST REGRESSOR (General)
    ========================== =#
 
-@with_kw_noshow mutable struct RobustRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct RobustRegressor <: MMI.Deterministic
     rho::RobustRho           = HuberRho(0.1)
     lambda::Real             = 1.0
     gamma::Real              = 0.0
@@ -85,7 +85,7 @@ descr(::Type{RobustRegressor}) = "Robust regression with objective ``∑ρ(Xθ -
    HUBER REGRESSOR
    =============== =#
 
-@with_kw_noshow mutable struct HuberRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct HuberRegressor <: MMI.Deterministic
     delta::Real              = 0.5
     lambda::Real             = 1.0
     gamma::Real              = 0.0
@@ -105,7 +105,7 @@ descr(::Type{HuberRegressor}) = "Robust regression with objective ``∑ρ(Xθ -
    QUANTILE REGRESSOR
    ================== =#
 
-@with_kw_noshow mutable struct QuantileRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct QuantileRegressor <: MMI.Deterministic
     delta::Real              = 0.5
     lambda::Real             = 1.0
     gamma::Real              = 0.0
@@ -126,7 +126,7 @@ descr(::Type{QuantileRegressor}) = "Robust regression with objective ``∑ρ(Xθ
    LEAST ABSOLUTE DEVIATION REGRESSOR
    ================================== =#
 
-@with_kw_noshow mutable struct LADRegressor <: MLJBase.Deterministic
+@with_kw_noshow mutable struct LADRegressor <: MMI.Deterministic
     lambda::Real             = 1.0
     gamma::Real              = 0.0
     penalty::SymStr          = :l2

test/runtests.jl

@@ -1,6 +1,7 @@
 using MLJLinearModels, Test, LinearAlgebra, Random
-import MLJBase
-DO_COMPARISONS = false; include("testutils.jl")
+using MLJBase # not MLJModelInterface, to mimick the full interface
+
+DO_COMPARISONS = true; include("testutils.jl")
 
 m("UTILS"); include("utils.jl")