Merge pull request #179 from ayush-1506/master

Add rng to resampling methods (cv and holdout)

src/resampling.jl

@@ -1,7 +1,7 @@
 ## RESAMPLING STRATEGIES
 
 abstract type ResamplingStrategy <: MLJType end
-
+import Random
 # resampling strategies are `==` if they have the same type and their
 # field values are `==`:
 function ==(s1::S, s2::S) where S<:ResamplingStrategy
@@ -15,21 +15,23 @@ end
 mutable struct Holdout <: ResamplingStrategy
     fraction_train::Float64
     shuffle::Bool
-    function Holdout(fraction_train, shuffle)
+    rng::Union{Int,AbstractRNG}
+    function Holdout(fraction_train, shuffle, rng)
         0 < fraction_train && fraction_train < 1 ||
             error("fraction_train must be between 0 and 1.")
-        return new(fraction_train, shuffle)
+        return new(fraction_train, shuffle, rng)
     end
 end
-Holdout(; fraction_train=0.7, shuffle=false) = Holdout(fraction_train, shuffle)
+Holdout(; fraction_train=0.7, shuffle=false, rng=Random.GLOBAL_RNG) = Holdout(fraction_train, shuffle, rng)
 show_as_constructed(::Type{<:Holdout}) = true
 
 mutable struct CV <: ResamplingStrategy
     nfolds::Int
     parallel::Bool
-    shuffle::Bool ## TODO: add seed/rng 
+    shuffle::Bool ## TODO: add seed/rng
+    rng::Union{Int,AbstractRNG}
 end
-CV(; nfolds=6, parallel=true, shuffle=false) = CV(nfolds, parallel, shuffle)
+CV(; nfolds=6, parallel=true, shuffle=false, rng=Random.GLOBAL_RNG) = CV(nfolds, parallel, shuffle, rng)
 MLJBase.show_as_constructed(::Type{<:CV}) = true
 
 
@@ -67,6 +69,12 @@ function evaluate!(mach::Machine, resampling::Holdout;
                    measure=nothing, operation=predict,
                    rows=nothing, force=false,verbosity=1)
 
+    if resampling.rng isa Integer
+        rng = MersenneTwister(resampling.rng)
+    else
+        rng = resampling.rng
+    end
+
     if measure == nothing
         _measures = default_measure(mach.model)
         if _measures == nothing
@@ -84,7 +92,7 @@ function evaluate!(mach::Machine, resampling::Holdout;
         rows == nothing ? eachindex(y) : rows
 
     train, test = partition(all, resampling.fraction_train,
-                            shuffle=resampling.shuffle)
+                            shuffle=resampling.shuffle, rng=rng)
     if verbosity > 0
             all == eachindex(y) ? "Resampling from all rows. " : "Resampling from a subset of all rows. "
         which_rows =
@@ -115,6 +123,12 @@ function evaluate!(mach::Machine, resampling::CV;
                    measure=nothing, operation=predict,
                    rows=nothing, force=false, verbosity=1)
 
+    if resampling.rng isa Integer
+        rng = MersenneTwister(resampling.rng)
+    else
+        rng = resampling.rng
+    end
+
     if measure == nothing
         _measures = default_measure(mach.model)
         if _measures == nothing
@@ -146,7 +160,7 @@ function evaluate!(mach::Machine, resampling::CV;
     nfolds = resampling.nfolds
 
     if resampling.shuffle
-        all = shuffle(all)
+        shuffle!(rng, collect(all))
     end
 
     k = floor(Int,n_samples/nfolds)