Removing GaussianProcesses

Project.toml

@@ -8,7 +8,6 @@ Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
 CellularAutomata = "878138dc-5b27-11ea-1a71-cb95d38d6b29"
 Distances = "b4f34e82-e78d-54a5-968a-f98e89d6e8f7"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
-GaussianProcesses = "891a1506-143c-57d2-908e-e1f8e92e6de9"
 LIBSVM = "b1bec4e5-fd48-53fe-b0cb-9723c09d164b"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 MLJLinearModels = "6ee0df7b-362f-4a72-a706-9e79364fb692"
@@ -22,7 +21,6 @@ Adapt = "3.3.3"
 CellularAutomata = "0.0.2"
 Distances = "0.10"
 Distributions = "0.24, 0.25"
-GaussianProcesses = "0.12"
 LIBSVM = "0.8"
 MLJLinearModels = "0.5, 0.6, 0.7"
 NNlib = "0.8.4"

src/ReservoirComputing.jl

@@ -4,7 +4,6 @@ using Adapt
 using CellularAutomata
 using Distances
 using Distributions
-using GaussianProcesses
 using LIBSVM
 using LinearAlgebra
 using MLJLinearModels
@@ -35,7 +34,6 @@ abstract type AbstractOutputLayer end
 abstract type AbstractPrediction end
 #training methods
 abstract type AbstractLinearModel end
-abstract type AbstractGaussianProcess end
 abstract type AbstractSupportVector end
 #should probably move some of these
 abstract type AbstractVariation end
@@ -81,7 +79,6 @@ include("predict.jl")
 
 #general training
 include("train/linear_regression.jl")
-include("train/gaussian_regression.jl")
 include("train/supportvector_regression.jl")
 
 #esn

src/predict.jl

@@ -46,20 +46,6 @@ function get_prediction(training_method::AbstractLinearModel, output_layer, x)
     return output_layer.output_matrix * x
 end
 
-#gaussian regression
-function get_prediction(training_method::AbstractGaussianProcess, output_layer, x)
-    out, sigma = zeros(output_layer.out_size), zeros(output_layer.out_size)
-
-    for j in 1:(output_layer.out_size)
-        x_new = reshape(x, length(x), 1)
-        gr = GaussianProcesses.predict_y(output_layer.output_matrix[j], x_new)
-        out[j] = gr[1][1]
-        sigma[j] = gr[2][1]
-    end
-
-    return (out, sigma)
-end
-
 #support vector regression
 function get_prediction(training_method::LIBSVM.AbstractSVR, output_layer, x)
     out = zeros(output_layer.out_size)
@@ -72,28 +58,11 @@ function get_prediction(training_method::LIBSVM.AbstractSVR, output_layer, x)
     return out
 end
 
-#creation of matrices for storing gaussian results (outs and sigmas)
-function output_storing(training_method::AbstractGaussianProcess,
-                        out_size,
-                        prediction_len,
-                        storing_type)
-    out = Adapt.adapt(storing_type, zeros(out_size, prediction_len))
-    sigma = Adapt.adapt(storing_type, zeros(out_size, prediction_len))
-    return (out, sigma)
-end
-
 #single matrix for other training methods
 function output_storing(training_method, out_size, prediction_len, storing_type)
     return Adapt.adapt(storing_type, zeros(out_size, prediction_len))
 end
 
-#storing results for gaussian training, getting also the sigmas
-function store_results!(training_method::AbstractGaussianProcess, out, output, i)
-    output[1][:, i] = out[1]
-    output[2][:, i] = out[2]
-    return out[1]
-end
-
 #general storing -> single matrix
 function store_results!(training_method, out, output, i)
     output[:, i] = out

test/esn/test_train.jl

@@ -1,4 +1,4 @@
-using ReservoirComputing, GaussianProcesses, MLJLinearModels, Random, Statistics, LIBSVM
+using ReservoirComputing, MLJLinearModels, Random, Statistics, LIBSVM
 
 const res_size = 20
 const ts = 0.0:0.1:50.0
@@ -18,17 +18,11 @@ training_methods = [
     StandardRidge(regularization_coeff = reg),
     LinearModel(RidgeRegression, regression_kwargs = (; lambda = reg)),
     LinearModel(regression = RidgeRegression, regression_kwargs = (; lambda = reg)),
-    GaussianProcess(MeanZero(), Poly(1.0, 1.0, 2)),
     EpsilonSVR(),
 ]
 
 for t in training_methods
     output_layer = train(esn, target_data, t)
     output = esn(Predictive(input_data), output_layer)
-    if t isa GaussianProcess
-        output = output[1]
-    else
-        output = output
-    end
     @test mean(abs.(target_data .- output)) ./ mean(abs.(target_data)) < 0.21
 end