Merge branch 'master' into compathelper/new_version/2020-03-03-00-12-18-450-1857031776

Author: xiaodaigh

NEWS.md

@@ -0,0 +1,3 @@
+# 0.1.1
+
+* Updated dependencies

Project.toml

@@ -1,7 +1,7 @@
 name = "JLBoostMLJ"
 uuid = "8b86df2c-1bc3-481d-95df-1c4d5a20ed95"
 authors = ["Dai ZJ <zhuojia.dai@gmail.com>"]
-version = "0.1.0"
+version = "0.1.1"
 
 [deps]
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
@@ -12,15 +12,16 @@ MLJBase = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 ScientificTypes = "321657f4-b219-11e9-178b-2701a2544e81"
 
 [compat]
-DataFrames = "0.19, 0.20"
-JLBoost = "^0.1.4"
+DataFrames = "0.20"
+JLBoost = "^0.1.7"
 LossFunctions = "0.5"
-MLJ = "0.6"
-MLJBase = "0.9, 0.11"
-ScientificTypes = "0.3"
+MLJ = "0.9"
+MLJBase = "0.11"
+ScientificTypes = "0.7"
 julia = "1"
 
 [extras]
+CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]

README.jl

@@ -1,4 +1,3 @@
-
 using RDatasets;
 iris = dataset("datasets", "iris");
 iris[!, :is_setosa] .= iris.Species .== "setosa";
@@ -52,4 +51,3 @@ predict(model, mljmodel.fitresult, X)
 
 
 feature_importance(mljmodel.fitresult.treemodel, X, y)
-

README.jmd

@@ -48,7 +48,6 @@ feature_importance(fitted_params(mljmachine).fitresult, X, y)
 Data preparation: need to convert `y` to categorical
 
 ```julia
-using CategoricalArrays
 y_cate = categorical(y)
 ```
 

README.md

@@ -19,15 +19,16 @@ model = JLBoostClassifier()
 
 
 ````
-JLBoostClassifier(loss = JLBoost.LogitLogLoss(),
-                  nrounds = 1,
-                  subsample = 1.0,
-                  eta = 1.0,
-                  max_depth = 6,
-                  min_child_weight = 1.0,
-                  lambda = 0.0,
-                  gamma = 0.0,
-                  colsample_bytree = 1,) @ 6…52
+JLBoostClassifier(
+    loss = JLBoost.LogitLogLoss(),
+    nrounds = 1,
+    subsample = 1.0,
+    eta = 1.0,
+    max_depth = 6,
+    min_child_weight = 1.0,
+    lambda = 0.0,
+    gamma = 0.0,
+    colsample_bytree = 1) @ 1…51
 ````
 
 
@@ -44,7 +45,7 @@ mljmachine  = machine(model, X, y)
 
 
 ````
-Machine{JLBoostClassifier} @ 1…93
+Machine{JLBoostClassifier} @ 3…89
 ````
 
 
@@ -73,7 +74,7 @@ Choosing a split on SepalLength
 Choosing a split on SepalWidth
 Choosing a split on PetalLength
 Choosing a split on PetalWidth
-Machine{JLBoostClassifier} @ 1…93
+Machine{JLBoostClassifier} @ 3…89
 ````
 
 
@@ -139,7 +140,6 @@ feature_importance(fitted_params(mljmachine).fitresult, X, y)
 Data preparation: need to convert `y` to categorical
 
 ````julia
-using CategoricalArrays
 y_cate = categorical(y)
 ````
 
@@ -184,10 +184,13 @@ r3 = range(jlb, :eta, lower=0.1, upper=1.0)
 
 
 ````
-MLJ.NumericRange(field = :eta,
-                 lower = 0.1,
-                 upper = 1.0,
-                 scale = :linear,) @ 1…03
+NumericRange(
+    field = :eta,
+    lower = 0.1,
+    upper = 1.0,
+    origin = 0.55,
+    unit = 0.45,
+    scale = :linear) @ 1…28
 ````
 
 
@@ -202,7 +205,7 @@ m = machine(tm, X, y_cate)
 
 
 ````
-Machine{ProbabilisticTunedModel} @ 5…78
+Machine{ProbabilisticTunedModel} @ 1…91
 ````
 
 
@@ -216,7 +219,7 @@ fit!(m)
 
 
 ````
-Machine{ProbabilisticTunedModel} @ 5…78
+Machine{ProbabilisticTunedModel} @ 1…91
 ````
 
 

build-readme.jl

@@ -1,10 +1,14 @@
 # Weave readme
 using Pkg
 Pkg.activate("c:/git/JLBoostMLJ")
+
+Pkg.add("Weave")
 using Weave
 
 weave("c:/git/JLBoostMLJ/README.jmd", out_path="c:/git/JLBoostMLJ", doctype="github")
 
+Pkg.rm("Weave")
+
 
 if false
 tangle("c:/git/JLBoostMLJ/README.jmd")

src/mlj.jl

@@ -8,7 +8,7 @@ import MLJBase: input_scitype, target_scitype, docstring, UnivariateFinite
 using ScientificTypes: Continuous, OrderedFactor, Count, Multiclass, Finite
 
 using LossFunctions: PoissonLoss, L2DistLoss
-using JLBoost: LogitLogLoss, jlboost, AUC, gini, feature_importance
+using JLBoost: LogitLogLoss, jlboost, AUC, gini, feature_importance, predict
 
 using DataFrames: DataFrame, nrow, levels, categorical
 
@@ -173,7 +173,7 @@ fit(model::JLBoostClassifier, verbosity::Int, X, y::AbstractVector) = begin
             fitresult = fitresult,
             cache = nothing,
             report = (
-                AUC = abs(AUC(predict(fitresult.treemodel, X), y)),
+                AUC = abs(AUC(JLBoost.predict(fitresult.treemodel, X), y)),
                 feature_importance = feature_importance(fitresult.treemodel, df)
             )
         )
@@ -195,15 +195,15 @@ fitted_params(model::JLBoostMLJModel, fitresult) = (fitresult = fitresult.treemo
 
 #  seehttps://alan-turing-institute.github.io/MLJ.jl/stable/adding_models_for_general_use/#The-predict-method-1
 predict(model::JLBoostClassifier, fitresult, Xnew) = begin
-    res = predict(fitresult.treemodel, Xnew)
+    res = JLBoost.predict(fitresult.treemodel, Xnew)
     p = 1 ./ (1 .+ exp.(-res))
     levels_cate = categorical(fitresult.target_levels)
     [UnivariateFinite(levels_cate, [p, 1-p]) for p in p]
 end
 
 
 predict(model::JLBoostMLJModel, fitresult, Xnew) = begin
-    predict(fitresult, Xnew)
+    JLBoost.predict(fitresult, Xnew)
 end
 
 # see https://alan-turing-institute.github.io/MLJ.jl/stable/adding_models_for_general_use/#Trait-declarations-1