Relax label vector type to AbstractVector (JuliaML#65)

* Relax label vector type to AbstractVector

* info for test

* update README

* add test cases for JLD2

Author: iblislin

.editorconfig

@@ -0,0 +1,3 @@
+[*.jl]
+indent_style = space
+indent_size = 4

Project.toml

@@ -15,8 +15,11 @@ julia = "1.3"
 
 [extras]
 DelimitedFiles = "8bb1440f-4735-579b-a4ab-409b98df4dab"
+FileIO = "5789e2e9-d7fb-5bc7-8068-2c6fae9b9549"
+JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819"
+RDatasets = "ce6b1742-4840-55fa-b093-852dadbb1d8b"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["DelimitedFiles", "Test", "SparseArrays"]
+test = ["DelimitedFiles", "FileIO", "JLD2", "RDatasets", "SparseArrays", "Test"]

README.md

@@ -22,26 +22,33 @@ for options.
 ```julia
 using LIBSVM
 using RDatasets
-using Printf, Statistics
+using Printf
+using Statistics
 
 # Load Fisher's classic iris data
 iris = dataset("datasets", "iris")
 
+# First four dimension of input data is features
+X = Matrix(iris[:, 1:4])'
+
 # LIBSVM handles multi-class data automatically using a one-against-one strategy
-labels = levelcode.(iris[:Species])
+y = iris.Species
 
-# First dimension of input data is features; second is instances
-instances = convert(Array, iris[:, 1:4])'
+# Split the dataset into training set and testing set
+Xtrain = X[:, 1:2:end]
+Xtest  = X[:, 2:2:end]
+ytrain = y[1:2:end]
+ytest  = y[2:2:end]
 
 # Train SVM on half of the data using default parameters. See documentation
 # of svmtrain for options
-model = svmtrain(instances[:, 1:2:end], labels[1:2:end]);
+model = svmtrain(Xtrain, ytrain)
 
 # Test model on the other half of the data.
-(predicted_labels, decision_values) = svmpredict(model, instances[:, 2:2:end]);
+ŷ, decision_values = svmpredict(model, Xtest);
 
 # Compute accuracy
-@printf "Accuracy: %.2f%%\n" mean((predicted_labels .== labels[2:2:end]))*100
+@printf "Accuracy: %.2f%%\n" mean(ŷ .== ytest) * 100
 ```
 
 ### ScikitLearn API
@@ -52,22 +59,29 @@ You can alternatively use `ScikitLearn.jl` API with same options as `svmtrain`:
 using LIBSVM
 using RDatasets
 
-#Classification C-SVM
+# Classification C-SVM
 iris = dataset("datasets", "iris")
-labels = levelcode.(iris[:, :Species])
-instances = convert(Array, iris[:, 1:4])
-model = fit!(SVC(), instances[1:2:end, :], labels[1:2:end])
-yp = predict(model, instances[2:2:end, :])
+X = Matrix(iris[:, 1:4])
+y = iris.Species
+
+Xtrain = X[1:2:end, :]
+Xtest  = X[2:2:end, :]
+ytrain = y[1:2:end]
+ytest  = y[2:2:end]
+
+model = fit!(SVC(), Xtrain, ytrain)
+ŷ = predict(model, Xtest)
+```
+
+```julia
+# Epsilon-Regression
 
-#epsilon-regression
 whiteside = RDatasets.dataset("MASS", "whiteside")
-X = Array(whiteside[:Gas])
-if typeof(X) <: AbstractVector
-    X = reshape(X, (length(X),1))
-end
-y = Array(whiteside[:Temp])
-svrmod = fit!(EpsilonSVR(cost = 10., gamma = 1.), X, y)
-yp = predict(svrmod, X)
+X = Matrix(whiteside[:, 3:3])  # the `Gas` column
+y = whiteside.Temp
+
+model = fit!(EpsilonSVR(cost = 10., gamma = 1.), X, y)
+ŷ = predict(model, X)
 ```
 
 ## Credits

src/LIBSVM.jl

@@ -1,22 +1,24 @@
-__precompile__()
 module LIBSVM
+
+
 import LIBLINEAR
+
 using SparseArrays
 using libsvm_jll
 
 export svmtrain, svmpredict, fit!, predict, transform,
-        SVC, NuSVC, OneClassSVM, NuSVR, EpsilonSVR, LinearSVC,
-        Linearsolver, Kernel
+       SVC, NuSVC, OneClassSVM, NuSVR, EpsilonSVR, LinearSVC,
+       Linearsolver, Kernel
 
 include("LibSVMtypes.jl")
 include("constants.jl")
 
 verbosity = false
 
-struct SupportVectors{T, U}
+struct SupportVectors{T,U}
     l::Int32
     nSV::Vector{Int32}
-    y::Vector{T}
+    y::AbstractVector{T}
     X::AbstractMatrix{U}
     indices::Vector{Int32}
     SVnodes::Vector{SVMNode}
@@ -36,8 +38,7 @@ function SupportVectors(smc::SVMModel, y, X)
 
     yi = smc.param.svm_type == 2 ? Float64[] : y[sv_indices]
 
-    SupportVectors(smc.l, nSV, yi , X[:,sv_indices],
-                        sv_indices, nodes)
+    SupportVectors(smc.l, nSV, yi , X[:,sv_indices], sv_indices, nodes)
 end
 
 struct SVM{T}
@@ -68,7 +69,7 @@ struct SVM{T}
     probability::Bool
 end
 
-function SVM(smc::SVMModel, y::T, X, weights, labels, svmtype, kernel) where T
+function SVM(smc::SVMModel, y, X, weights, labels, svmtype, kernel)
     svs = SupportVectors(smc, y, X)
     coefs = zeros(smc.l, smc.nr_class-1)
     for k in 1:(smc.nr_class-1)
@@ -266,39 +267,48 @@ function set_num_threads(nt::Integer)
 end
 
 """
-```julia
-svmtrain{T, U<:Real}(X::AbstractMatrix{U}, y::AbstractVector{T}=[];
-    svmtype::Type=SVC, kernel::Kernel.KERNEL=Kernel.RadialBasis, degree::Integer=3,
-    gamma::Float64=1.0/size(X, 1), coef0::Float64=0.0,
-    cost::Float64=1.0, nu::Float64=0.5, epsilon::Float64=0.1,
-    tolerance::Float64=0.001, shrinking::Bool=true,
-    probability::Bool=false, weights::Union{Dict{T, Float64}, Cvoid}=nothing,
-    cachesize::Float64=200.0, verbose::Bool=false)
-```
+    svmtrain(
+        X::AbstractMatrix{U}, y::AbstractVector{T} = [];
+        svmtype::Type = SVC,
+        kernel::Kernel.KERNEL = Kernel.RadialBasis,
+        degree::Integer = 3,
+        gamma::Float64 = 1.0/size(X, 1),
+        coef0::Float64 = 0.0,
+        cost::Float64=1.0,
+        nu::Float64 = 0.5,
+        epsilon::Float64 = 0.1,
+        tolerance::Float64 = 0.001,
+        shrinking::Bool = true,
+        probability::Bool = false,
+        weights::Union{Dict{T,Float64},Cvoid} = nothing,
+        cachesize::Float64 = 200.0,
+        verbose::Bool = false
+    ) where {T,U<:Real}
+
 Train Support Vector Machine using LIBSVM using response vector `y`
-and training data `X`. The shape of `X` needs to be (nfeatures, nsamples).
+and training data `X`. The shape of `X` needs to be `(nfeatures, nsamples)`.
 For one-class SVM use only `X`.
 
 # Arguments
 
-* `svmtype::Type=LIBSVM.SVC`: Type of SVM to train `SVC` (for C-SVM), `NuSVC`
+* `svmtype::Type = LIBSVM.SVC`: Type of SVM to train `SVC` (for C-SVM), `NuSVC`
     `OneClassSVM`, `EpsilonSVR` or `NuSVR`. Defaults to `OneClassSVM` if
     `y` is not used.
-* `kernel::Kernels.KERNEL=Kernel.RadialBasis`: Model kernel `Linear`, `Polynomial`,
+* `kernel::Kernels.KERNEL = Kernel.RadialBasis`: Model kernel `Linear`, `Polynomial`,
     `RadialBasis`, `Sigmoid` or `Precomputed`.
-* `degree::Integer=3`: Kernel degree. Used for polynomial kernel
-* `gamma::Float64=1.0/size(X, 1)` : γ for kernels
-* `coef0::Float64=0.0`: parameter for sigmoid and polynomial kernel
-* `cost::Float64=1.0`: cost parameter C of C-SVC, epsilon-SVR, and nu-SVR
-* `nu::Float64=0.5`: parameter nu of nu-SVC, one-class SVM, and nu-SVR
-* `epsilon::Float64=0.1`: epsilon in loss function of epsilon-SVR
-* `tolerance::Float64=0.001`: tolerance of termination criterion
-* `shrinking::Bool=true`: whether to use the shrinking heuristics
-* `probability::Bool=false`: whether to train a SVC or SVR model for probability estimates
+* `degree::Integer = 3`: Kernel degree. Used for polynomial kernel
+* `gamma::Float64 = 1.0/size(X, 1)` : γ for kernels
+* `coef0::Float64 = 0.0`: parameter for sigmoid and polynomial kernel
+* `cost::Float64 = 1.0`: cost parameter C of C-SVC, epsilon-SVR, and nu-SVR
+* `nu::Float64 = 0.5`: parameter nu of nu-SVC, one-class SVM, and nu-SVR
+* `epsilon::Float64 = 0.1`: epsilon in loss function of epsilon-SVR
+* `tolerance::Float64 = 0.001`: tolerance of termination criterion
+* `shrinking::Bool = true`: whether to use the shrinking heuristics
+* `probability::Bool = false`: whether to train a SVC or SVR model for probability estimates
 * `weights::Union{Dict{T, Float64}, Cvoid}=nothing`: dictionary of class weights
-* `cachesize::Float64=100.0`: cache memory size in MB
-* `verbose::Bool=false`: print training output from LIBSVM if true
-* `nt::Integer=0`: number of OpenMP cores to use, if 0 it is set to OMP_NUM_THREADS, if negative it is set to the max number of threads
+* `cachesize::Float64 = 100.0`: cache memory size in MB
+* `verbose::Bool = false`: print training output from LIBSVM if true
+* `nt::Integer = 0`: number of OpenMP cores to use, if 0 it is set to OMP_NUM_THREADS, if negative it is set to the max number of threads
 
 Consult LIBSVM documentation for advice on the choise of correct
 parameters and model tuning.
@@ -419,4 +429,5 @@ end
 include("ScikitLearnTypes.jl")
 include("ScikitLearnAPI.jl")
 
+
 end

src/ScikitLearnAPI.jl

@@ -65,7 +65,7 @@ LinearSVC(;solver = Linearsolver.L2R_L2LOSS_SVC_DUAL,
           cost, p, bias, verbose, nothing)
 @declare_hyperparameters(LinearSVC, [:solver, :weights, :tolerance, :cost, :p, :bias])
 
-function fit!(model::Union{AbstractSVC,AbstractSVR}, X::AbstractMatrix, y::Vector=[])
+function fit!(model::Union{AbstractSVC,AbstractSVR}, X::AbstractMatrix, y::AbstractVector = [])
     #Build arguments for calling svmtrain
     model.gamma == :auto && (model.gamma = 1.0/size(X', 1))
     kwargs = Tuple{Symbol, Any}[]
@@ -97,7 +97,7 @@ function get_params(model::Union{AbstractSVC,AbstractSVR, LinearSVC})
     return params
 end
 
-function fit!(model::LinearSVC, X::AbstractMatrix, y::Vector)
+function fit!(model::LinearSVC, X::AbstractMatrix, y::AbstractVector)
     model.fit = LIBLINEAR.linear_train(y, X', solver_type = Int32(model.solver),
     weights = model.weights, C = model.cost, bias = model.bias,
     p = model.p, eps = model.tolerance, verbose = model.verbose)