rumale 0.18.6 → 0.19.3
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/.rubocop.yml +80 -3
- data/CHANGELOG.md +45 -0
- data/Gemfile +2 -0
- data/README.md +5 -36
- data/lib/rumale.rb +5 -0
- data/lib/rumale/base/base_estimator.rb +2 -0
- data/lib/rumale/clustering/dbscan.rb +4 -0
- data/lib/rumale/clustering/gaussian_mixture.rb +2 -0
- data/lib/rumale/clustering/hdbscan.rb +3 -1
- data/lib/rumale/clustering/k_means.rb +2 -1
- data/lib/rumale/clustering/k_medoids.rb +5 -1
- data/lib/rumale/clustering/mini_batch_k_means.rb +139 -0
- data/lib/rumale/clustering/power_iteration.rb +2 -0
- data/lib/rumale/clustering/single_linkage.rb +2 -0
- data/lib/rumale/dataset.rb +5 -3
- data/lib/rumale/decomposition/factor_analysis.rb +2 -0
- data/lib/rumale/decomposition/pca.rb +24 -5
- data/lib/rumale/ensemble/ada_boost_classifier.rb +3 -0
- data/lib/rumale/ensemble/ada_boost_regressor.rb +3 -0
- data/lib/rumale/evaluation_measure/function.rb +2 -1
- data/lib/rumale/evaluation_measure/normalized_mutual_information.rb +2 -0
- data/lib/rumale/evaluation_measure/precision_recall.rb +5 -0
- data/lib/rumale/evaluation_measure/roc_auc.rb +3 -0
- data/lib/rumale/evaluation_measure/silhouette_score.rb +2 -0
- data/lib/rumale/feature_extraction/feature_hasher.rb +14 -1
- data/lib/rumale/feature_extraction/hash_vectorizer.rb +1 -0
- data/lib/rumale/feature_extraction/tfidf_transformer.rb +113 -0
- data/lib/rumale/kernel_approximation/nystroem.rb +1 -1
- data/lib/rumale/kernel_machine/kernel_ridge.rb +2 -0
- data/lib/rumale/kernel_machine/kernel_svc.rb +1 -1
- data/lib/rumale/linear_model/base_linear_model.rb +3 -1
- data/lib/rumale/linear_model/base_sgd.rb +1 -1
- data/lib/rumale/linear_model/linear_regression.rb +1 -0
- data/lib/rumale/linear_model/ridge.rb +1 -0
- data/lib/rumale/manifold/mds.rb +2 -0
- data/lib/rumale/manifold/tsne.rb +4 -0
- data/lib/rumale/metric_learning/neighbourhood_component_analysis.rb +14 -1
- data/lib/rumale/model_selection/cross_validation.rb +3 -2
- data/lib/rumale/model_selection/grid_search_cv.rb +1 -0
- data/lib/rumale/model_selection/k_fold.rb +1 -1
- data/lib/rumale/model_selection/shuffle_split.rb +1 -1
- data/lib/rumale/multiclass/one_vs_rest_classifier.rb +2 -2
- data/lib/rumale/nearest_neighbors/k_neighbors_classifier.rb +1 -0
- data/lib/rumale/nearest_neighbors/k_neighbors_regressor.rb +2 -0
- data/lib/rumale/nearest_neighbors/vp_tree.rb +1 -1
- data/lib/rumale/neural_network/adam.rb +1 -1
- data/lib/rumale/neural_network/base_mlp.rb +2 -1
- data/lib/rumale/optimizer/ada_grad.rb +3 -0
- data/lib/rumale/optimizer/adam.rb +3 -0
- data/lib/rumale/optimizer/nadam.rb +5 -0
- data/lib/rumale/optimizer/rmsprop.rb +3 -0
- data/lib/rumale/optimizer/sgd.rb +3 -0
- data/lib/rumale/optimizer/yellow_fin.rb +3 -0
- data/lib/rumale/pipeline/pipeline.rb +3 -0
- data/lib/rumale/polynomial_model/base_factorization_machine.rb +6 -1
- data/lib/rumale/polynomial_model/factorization_machine_classifier.rb +5 -0
- data/lib/rumale/polynomial_model/factorization_machine_regressor.rb +5 -0
- data/lib/rumale/preprocessing/binarizer.rb +60 -0
- data/lib/rumale/preprocessing/l1_normalizer.rb +62 -0
- data/lib/rumale/preprocessing/l2_normalizer.rb +2 -1
- data/lib/rumale/preprocessing/max_normalizer.rb +62 -0
- data/lib/rumale/preprocessing/one_hot_encoder.rb +3 -0
- data/lib/rumale/preprocessing/ordinal_encoder.rb +2 -0
- data/lib/rumale/preprocessing/polynomial_features.rb +1 -0
- data/lib/rumale/probabilistic_output.rb +2 -0
- data/lib/rumale/tree/base_decision_tree.rb +2 -0
- data/lib/rumale/tree/decision_tree_classifier.rb +1 -0
- data/lib/rumale/tree/gradient_tree_regressor.rb +1 -0
- data/lib/rumale/utils.rb +1 -0
- data/lib/rumale/validation.rb +7 -0
- data/lib/rumale/version.rb +1 -1
- data/rumale.gemspec +1 -3
- metadata +11 -34
@@ -69,7 +69,7 @@ module Rumale
|
|
69
69
|
n_components = [1, [@params[:n_components], n_samples].min].max
|
70
70
|
|
71
71
|
# random sampling.
|
72
|
-
@component_indices = Numo::Int32.cast(
|
72
|
+
@component_indices = Numo::Int32.cast(Array(0...n_samples).shuffle(random: sub_rng)[0...n_components])
|
73
73
|
@components = x[@component_indices, true]
|
74
74
|
|
75
75
|
# calculate normalizing factor.
|
@@ -30,6 +30,7 @@ module Rumale
|
|
30
30
|
def initialize(reg_param: 1.0)
|
31
31
|
raise TypeError, 'Expect class of reg_param to be Float or Numo::DFloat' unless reg_param.is_a?(Float) || reg_param.is_a?(Numo::DFloat)
|
32
32
|
raise ArgumentError, 'Expect reg_param array to be 1-D arrray' if reg_param.is_a?(Numo::DFloat) && reg_param.shape.size != 1
|
33
|
+
|
33
34
|
@params = {}
|
34
35
|
@params[:reg_param] = reg_param
|
35
36
|
@weight_vec = nil
|
@@ -55,6 +56,7 @@ module Rumale
|
|
55
56
|
@weight_vec = Numo::Linalg.solve(reg_kernel_mat, y, driver: 'sym')
|
56
57
|
else
|
57
58
|
raise ArgumentError, 'Expect y and reg_param to have the same number of elements.' unless y.shape[1] == @params[:reg_param].shape[0]
|
59
|
+
|
58
60
|
n_outputs = y.shape[1]
|
59
61
|
@weight_vec = Numo::DFloat.zeros(n_samples, n_outputs)
|
60
62
|
n_outputs.times do |n|
|
@@ -172,7 +172,7 @@ module Rumale
|
|
172
172
|
# Start optimization.
|
173
173
|
@params[:max_iter].times do |t|
|
174
174
|
# random sampling
|
175
|
-
rand_ids =
|
175
|
+
rand_ids = Array(0...n_training_samples).shuffle(random: sub_rng) if rand_ids.empty?
|
176
176
|
target_id = rand_ids.shift
|
177
177
|
# update the weight vector
|
178
178
|
func = (weight_vec * bin_y).dot(x[target_id, true].transpose).to_f
|
@@ -8,6 +8,7 @@ module Rumale
|
|
8
8
|
# @note
|
9
9
|
# In version 0.17.0, a new linear model abstract class called BaseSGD is introduced.
|
10
10
|
# BaseLienarModel is deprecated and will be removed in the future.
|
11
|
+
# @deprecated Use BaseSGD class instead. This class will be deleted in version 0.20.0.
|
11
12
|
#
|
12
13
|
# BaseLinearModel is an abstract class for implementation of linear estimator
|
13
14
|
# with mini-batch stochastic gradient descent optimization.
|
@@ -55,7 +56,7 @@ module Rumale
|
|
55
56
|
samples = @params[:fit_bias] ? expand_feature(x) : x
|
56
57
|
# Initialize some variables.
|
57
58
|
n_samples, n_features = samples.shape
|
58
|
-
rand_ids =
|
59
|
+
rand_ids = Array(0...n_samples).shuffle(random: @rng.dup)
|
59
60
|
weight = Numo::DFloat.zeros(n_features)
|
60
61
|
optimizer = @params[:optimizer].dup
|
61
62
|
# Optimization.
|
@@ -68,6 +69,7 @@ module Rumale
|
|
68
69
|
# Update weight.
|
69
70
|
loss_gradient = calc_loss_gradient(sub_samples, sub_targets, weight)
|
70
71
|
next if loss_gradient.ne(0.0).count.zero?
|
72
|
+
|
71
73
|
weight = calc_new_weight(optimizer, sub_samples, weight, loss_gradient)
|
72
74
|
end
|
73
75
|
split_weight(weight)
|
@@ -209,7 +209,7 @@ module Rumale
|
|
209
209
|
l1_penalty = LinearModel::Penalty::L1Penalty.new(reg_param: l1_reg_param) if apply_l1_penalty?
|
210
210
|
# Optimization.
|
211
211
|
@params[:max_iter].times do |t|
|
212
|
-
sample_ids =
|
212
|
+
sample_ids = Array(0...n_samples)
|
213
213
|
sample_ids.shuffle!(random: sub_rng)
|
214
214
|
until (subset_ids = sample_ids.shift(@params[:batch_size])).empty?
|
215
215
|
# sampling
|
data/lib/rumale/manifold/mds.rb
CHANGED
@@ -83,6 +83,7 @@ module Rumale
|
|
83
83
|
def fit(x, _not_used = nil)
|
84
84
|
x = check_convert_sample_array(x)
|
85
85
|
raise ArgumentError, 'Expect the input distance matrix to be square.' if @params[:metric] == 'precomputed' && x.shape[0] != x.shape[1]
|
86
|
+
|
86
87
|
# initialize some varibales.
|
87
88
|
n_samples = x.shape[0]
|
88
89
|
hi_distance_mat = @params[:metric] == 'precomputed' ? x : Rumale::PairwiseMetric.euclidean_distance(x)
|
@@ -142,6 +143,7 @@ module Rumale
|
|
142
143
|
def terminate?(old_stress, new_stress)
|
143
144
|
return false if @params[:tol].nil?
|
144
145
|
return false if old_stress.nil?
|
146
|
+
|
145
147
|
(old_stress - new_stress).abs <= @params[:tol]
|
146
148
|
end
|
147
149
|
|
data/lib/rumale/manifold/tsne.rb
CHANGED
@@ -89,6 +89,7 @@ module Rumale
|
|
89
89
|
def fit(x, _not_used = nil)
|
90
90
|
x = check_convert_sample_array(x)
|
91
91
|
raise ArgumentError, 'Expect the input distance matrix to be square.' if @params[:metric] == 'precomputed' && x.shape[0] != x.shape[1]
|
92
|
+
|
92
93
|
# initialize some varibales.
|
93
94
|
@n_iter = 0
|
94
95
|
distance_mat = @params[:metric] == 'precomputed' ? x**2 : Rumale::PairwiseMetric.squared_error(x)
|
@@ -99,6 +100,7 @@ module Rumale
|
|
99
100
|
one_vec = Numo::DFloat.ones(x.shape[0]).expand_dims(1)
|
100
101
|
@params[:max_iter].times do |t|
|
101
102
|
break if terminate?(hi_prob_mat, lo_prob_mat)
|
103
|
+
|
102
104
|
a = hi_prob_mat * lo_prob_mat
|
103
105
|
b = lo_prob_mat * lo_prob_mat
|
104
106
|
y = (b.dot(one_vec) * y + (a - b).dot(y)) / a.dot(one_vec)
|
@@ -170,6 +172,7 @@ module Rumale
|
|
170
172
|
entropy, probs = gaussian_distributed_probability_vector(sample_id, distance_vec, beta)
|
171
173
|
diff_entropy = entropy - init_entropy
|
172
174
|
break if diff_entropy.abs <= 1e-5
|
175
|
+
|
173
176
|
if diff_entropy.positive?
|
174
177
|
betamin = beta
|
175
178
|
if betamax == Float::MAX
|
@@ -211,6 +214,7 @@ module Rumale
|
|
211
214
|
|
212
215
|
def terminate?(p, q)
|
213
216
|
return false if @params[:tol].nil?
|
217
|
+
|
214
218
|
cost(p, q) <= @params[:tol]
|
215
219
|
end
|
216
220
|
end
|
@@ -2,13 +2,15 @@
|
|
2
2
|
|
3
3
|
require 'rumale/base/base_estimator'
|
4
4
|
require 'rumale/base/transformer'
|
5
|
-
require 'mopti/scaled_conjugate_gradient'
|
6
5
|
|
7
6
|
module Rumale
|
8
7
|
module MetricLearning
|
9
8
|
# NeighbourhoodComponentAnalysis is a class that implements Neighbourhood Component Analysis.
|
10
9
|
#
|
11
10
|
# @example
|
11
|
+
# require 'mopti'
|
12
|
+
# require 'rumale'
|
13
|
+
#
|
12
14
|
# transformer = Rumale::MetricLearning::NeighbourhoodComponentAnalysis.new
|
13
15
|
# transformer.fit(training_samples, traininig_labels)
|
14
16
|
# low_samples = transformer.transform(testing_samples)
|
@@ -63,6 +65,8 @@ module Rumale
|
|
63
65
|
# @param y [Numo::Int32] (shape: [n_samples]) The labels to be used for fitting the model.
|
64
66
|
# @return [NeighbourhoodComponentAnalysis] The learned classifier itself.
|
65
67
|
def fit(x, y)
|
68
|
+
raise 'NeighbourhoodComponentAnalysis#fit requires Mopti but that is not loaded.' unless enable_mopti?
|
69
|
+
|
66
70
|
x = check_convert_sample_array(x)
|
67
71
|
y = check_convert_label_array(y)
|
68
72
|
check_sample_label_size(x, y)
|
@@ -98,6 +102,14 @@ module Rumale
|
|
98
102
|
|
99
103
|
private
|
100
104
|
|
105
|
+
def enable_mopti?
|
106
|
+
if defined?(Mopti).nil?
|
107
|
+
warn('NeighbourhoodComponentAnalysis#fit requires Mopti but that is not loaded. You should intall and load mopti gem in advance.')
|
108
|
+
return false
|
109
|
+
end
|
110
|
+
true
|
111
|
+
end
|
112
|
+
|
101
113
|
def init_components(x, n_features, n_components)
|
102
114
|
if @params[:init] == 'pca'
|
103
115
|
pca = Rumale::Decomposition::PCA.new(n_components: n_components, solver: 'evd')
|
@@ -126,6 +138,7 @@ module Rumale
|
|
126
138
|
res = prm
|
127
139
|
puts "[NeighbourhoodComponentAnalysis] The value of objective function after #{res[:n_iter]} epochs: #{x.shape[0] - res[:fnc]}" if @params[:verbose]
|
128
140
|
break if (fold - res[:fnc]).abs <= @params[:tol] && (dold - res[:jcb]).abs <= @params[:tol]
|
141
|
+
|
129
142
|
fold = res[:fnc]
|
130
143
|
dold = res[:jcb]
|
131
144
|
end
|
@@ -69,10 +69,11 @@ module Rumale
|
|
69
69
|
# the return_train_score is false.
|
70
70
|
def perform(x, y)
|
71
71
|
x = check_convert_sample_array(x)
|
72
|
-
|
72
|
+
case @estimator
|
73
|
+
when Rumale::Base::Classifier
|
73
74
|
y = check_convert_label_array(y)
|
74
75
|
check_sample_label_size(x, y)
|
75
|
-
|
76
|
+
when Rumale::Base::Regressor
|
76
77
|
y = check_convert_tvalue_array(y)
|
77
78
|
check_sample_tvalue_size(x, y)
|
78
79
|
else
|
@@ -156,6 +156,7 @@ module Rumale
|
|
156
156
|
|
157
157
|
def valid_param_grid(grid)
|
158
158
|
raise TypeError, 'Expect class of param_grid to be Hash or Array' unless grid.is_a?(Hash) || grid.is_a?(Array)
|
159
|
+
|
159
160
|
grid = [grid] if grid.is_a?(Hash)
|
160
161
|
grid.each do |h|
|
161
162
|
raise TypeError, 'Expect class of elements in param_grid to be Hash' unless h.is_a?(Hash)
|
@@ -62,7 +62,7 @@ module Rumale
|
|
62
62
|
end
|
63
63
|
sub_rng = @rng.dup
|
64
64
|
# Splits dataset ids to each fold.
|
65
|
-
dataset_ids =
|
65
|
+
dataset_ids = Array(0...n_samples)
|
66
66
|
dataset_ids.shuffle!(random: sub_rng) if @shuffle
|
67
67
|
fold_sets = Array.new(@n_splits) do |n|
|
68
68
|
n_fold_samples = n_samples / @n_splits
|
@@ -74,7 +74,7 @@ module Rumale
|
|
74
74
|
end
|
75
75
|
sub_rng = @rng.dup
|
76
76
|
# Returns array consisting of the training and testing ids for each fold.
|
77
|
-
dataset_ids =
|
77
|
+
dataset_ids = Array(0...n_samples)
|
78
78
|
Array.new(@n_splits) do
|
79
79
|
test_ids = dataset_ids.sample(n_test_samples, random: sub_rng)
|
80
80
|
train_ids = if @train_size.nil?
|
@@ -1,7 +1,7 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
|
3
|
-
require 'rumale/base/base_estimator
|
4
|
-
require 'rumale/base/classifier
|
3
|
+
require 'rumale/base/base_estimator'
|
4
|
+
require 'rumale/base/classifier'
|
5
5
|
|
6
6
|
module Rumale
|
7
7
|
# This module consists of the classes that implement multi-class classification strategy.
|
@@ -67,6 +67,7 @@ module Rumale
|
|
67
67
|
y = check_convert_label_array(y)
|
68
68
|
check_sample_label_size(x, y)
|
69
69
|
raise ArgumentError, 'Expect the input distance matrix to be square.' if @params[:metric] == 'precomputed' && x.shape[0] != x.shape[1]
|
70
|
+
|
70
71
|
@prototypes = if @params[:metric] == 'euclidean'
|
71
72
|
if @params[:algorithm] == 'vptree'
|
72
73
|
VPTree.new(x)
|
@@ -61,6 +61,7 @@ module Rumale
|
|
61
61
|
y = check_convert_tvalue_array(y)
|
62
62
|
check_sample_tvalue_size(x, y)
|
63
63
|
raise ArgumentError, 'Expect the input distance matrix to be square.' if @params[:metric] == 'precomputed' && x.shape[0] != x.shape[1]
|
64
|
+
|
64
65
|
@prototypes = if @params[:metric] == 'euclidean'
|
65
66
|
if @params[:algorithm] == 'vptree'
|
66
67
|
VPTree.new(x)
|
@@ -82,6 +83,7 @@ module Rumale
|
|
82
83
|
if @params[:metric] == 'precomputed' && x.shape[1] != @values.shape[0]
|
83
84
|
raise ArgumentError, 'Expect the size input matrix to be n_testing_samples-by-n_training_samples.'
|
84
85
|
end
|
86
|
+
|
85
87
|
# Initialize some variables.
|
86
88
|
n_samples = x.shape[0]
|
87
89
|
n_prototypes, n_outputs = @values.shape
|
@@ -30,7 +30,7 @@ module Rumale
|
|
30
30
|
@params = {}
|
31
31
|
@params[:min_samples_leaf] = min_samples_leaf
|
32
32
|
@data = x
|
33
|
-
@tree = build_tree(Numo::Int32.cast(
|
33
|
+
@tree = build_tree(Numo::Int32.cast(Array(0...@data.shape[0])))
|
34
34
|
end
|
35
35
|
|
36
36
|
# Search k-nearest neighbors of given query point.
|
@@ -32,7 +32,7 @@ module Rumale
|
|
32
32
|
end
|
33
33
|
|
34
34
|
# @!visibility private
|
35
|
-
# Calculate the updated weight with
|
35
|
+
# Calculate the updated weight with Adam adaptive learning rate.
|
36
36
|
#
|
37
37
|
# @param weight [Numo::DFloat] (shape: [n_features]) The weight to be updated.
|
38
38
|
# @param gradient [Numo::DFloat] (shape: [n_features]) The gradient for updating the weight.
|
@@ -222,7 +222,7 @@ module Rumale
|
|
222
222
|
n_samples = x.shape[0]
|
223
223
|
|
224
224
|
@params[:max_iter].times do |t|
|
225
|
-
sample_ids =
|
225
|
+
sample_ids = Array(0...n_samples)
|
226
226
|
sample_ids.shuffle!(random: srng)
|
227
227
|
until (subset_ids = sample_ids.shift(@params[:batch_size])).empty?
|
228
228
|
# random sampling
|
@@ -233,6 +233,7 @@ module Rumale
|
|
233
233
|
# calc loss function
|
234
234
|
loss, dout = loss_func.call(out, sub_y)
|
235
235
|
break if loss < @params[:tol]
|
236
|
+
|
236
237
|
# backward
|
237
238
|
backward.call(dout)
|
238
239
|
end
|
@@ -7,6 +7,8 @@ module Rumale
|
|
7
7
|
module Optimizer
|
8
8
|
# AdaGrad is a class that implements AdaGrad optimizer.
|
9
9
|
#
|
10
|
+
# @deprecated AdaGrad will be deleted in version 0.20.0.
|
11
|
+
#
|
10
12
|
# *Reference*
|
11
13
|
# - Duchi, J., Hazan, E., and Singer, Y., "Adaptive Subgradient Methods for Online Learning and Stochastic Optimization," J. Machine Learning Research, vol. 12, pp. 2121--2159, 2011.
|
12
14
|
class AdaGrad
|
@@ -17,6 +19,7 @@ module Rumale
|
|
17
19
|
#
|
18
20
|
# @param learning_rate [Float] The initial value of learning rate.
|
19
21
|
def initialize(learning_rate: 0.01)
|
22
|
+
warn 'warning: AdaGrad is deprecated. This class will be deleted in version 0.20.0.'
|
20
23
|
check_params_numeric(learning_rate: learning_rate)
|
21
24
|
check_params_positive(learning_rate: learning_rate)
|
22
25
|
@params = {}
|
@@ -7,6 +7,8 @@ module Rumale
|
|
7
7
|
module Optimizer
|
8
8
|
# Adam is a class that implements Adam optimizer.
|
9
9
|
#
|
10
|
+
# @deprecated Adam will be deleted in version 0.20.0.
|
11
|
+
#
|
10
12
|
# *Reference*
|
11
13
|
# - Kingma, D P., and Ba, J., "Adam: A Method for Stochastic Optimization," Proc. ICLR'15, 2015.
|
12
14
|
class Adam
|
@@ -19,6 +21,7 @@ module Rumale
|
|
19
21
|
# @param decay1 [Float] The smoothing parameter for the first moment.
|
20
22
|
# @param decay2 [Float] The smoothing parameter for the second moment.
|
21
23
|
def initialize(learning_rate: 0.001, decay1: 0.9, decay2: 0.999)
|
24
|
+
warn 'warning: Adam is deprecated. This class will be deleted in version 0.20.0.'
|
22
25
|
check_params_numeric(learning_rate: learning_rate, decay1: decay1, decay2: decay2)
|
23
26
|
check_params_positive(learning_rate: learning_rate, decay1: decay1, decay2: decay2)
|
24
27
|
@params = {}
|
@@ -5,9 +5,13 @@ require 'rumale/base/base_estimator'
|
|
5
5
|
|
6
6
|
module Rumale
|
7
7
|
# This module consists of the classes that implement optimizers adaptively tuning hyperparameters.
|
8
|
+
#
|
9
|
+
# @deprecated Optimizer module will be deleted in version 0.20.0.
|
8
10
|
module Optimizer
|
9
11
|
# Nadam is a class that implements Nadam optimizer.
|
10
12
|
#
|
13
|
+
# @deprecated Nadam will be deleted in version 0.20.0.
|
14
|
+
#
|
11
15
|
# *Reference*
|
12
16
|
# - Dozat, T., "Incorporating Nesterov Momentum into Adam," Tech. Repo. Stanford University, 2015.
|
13
17
|
class Nadam
|
@@ -20,6 +24,7 @@ module Rumale
|
|
20
24
|
# @param decay1 [Float] The smoothing parameter for the first moment.
|
21
25
|
# @param decay2 [Float] The smoothing parameter for the second moment.
|
22
26
|
def initialize(learning_rate: 0.01, decay1: 0.9, decay2: 0.999)
|
27
|
+
warn 'warning: Nadam is deprecated. This class will be deleted in version 0.20.0.'
|
23
28
|
check_params_numeric(learning_rate: learning_rate, decay1: decay1, decay2: decay2)
|
24
29
|
check_params_positive(learning_rate: learning_rate, decay1: decay1, decay2: decay2)
|
25
30
|
@params = {}
|
@@ -7,6 +7,8 @@ module Rumale
|
|
7
7
|
module Optimizer
|
8
8
|
# RMSProp is a class that implements RMSProp optimizer.
|
9
9
|
#
|
10
|
+
# @deprecated RMSProp will be deleted in version 0.20.0.
|
11
|
+
#
|
10
12
|
# *Reference*
|
11
13
|
# - Sutskever, I., Martens, J., Dahl, G., and Hinton, G., "On the importance of initialization and momentum in deep learning," Proc. ICML' 13, pp. 1139--1147, 2013.
|
12
14
|
# - Hinton, G., Srivastava, N., and Swersky, K., "Lecture 6e rmsprop," Neural Networks for Machine Learning, 2012.
|
@@ -20,6 +22,7 @@ module Rumale
|
|
20
22
|
# @param momentum [Float] The initial value of momentum.
|
21
23
|
# @param decay [Float] The smooting parameter.
|
22
24
|
def initialize(learning_rate: 0.01, momentum: 0.9, decay: 0.9)
|
25
|
+
warn 'warning: RMSProp is deprecated. This class will be deleted in version 0.20.0.'
|
23
26
|
check_params_numeric(learning_rate: learning_rate, momentum: momentum, decay: decay)
|
24
27
|
check_params_positive(learning_rate: learning_rate, momentum: momentum, decay: decay)
|
25
28
|
@params = {}
|
data/lib/rumale/optimizer/sgd.rb
CHANGED
@@ -6,6 +6,8 @@ require 'rumale/base/base_estimator'
|
|
6
6
|
module Rumale
|
7
7
|
module Optimizer
|
8
8
|
# SGD is a class that implements SGD optimizer.
|
9
|
+
#
|
10
|
+
# @deprecated SGD will be deleted in version 0.20.0.
|
9
11
|
class SGD
|
10
12
|
include Base::BaseEstimator
|
11
13
|
include Validation
|
@@ -16,6 +18,7 @@ module Rumale
|
|
16
18
|
# @param momentum [Float] The initial value of momentum.
|
17
19
|
# @param decay [Float] The smooting parameter.
|
18
20
|
def initialize(learning_rate: 0.01, momentum: 0.0, decay: 0.0)
|
21
|
+
warn 'warning: SGD is deprecated. This class will be deleted in version 0.20.0.'
|
19
22
|
check_params_numeric(learning_rate: learning_rate, momentum: momentum, decay: decay)
|
20
23
|
check_params_positive(learning_rate: learning_rate, momentum: momentum, decay: decay)
|
21
24
|
@params = {}
|
@@ -7,6 +7,8 @@ module Rumale
|
|
7
7
|
module Optimizer
|
8
8
|
# YellowFin is a class that implements YellowFin optimizer.
|
9
9
|
#
|
10
|
+
# @deprecated YellowFin will be deleted in version 0.20.0.
|
11
|
+
#
|
10
12
|
# *Reference*
|
11
13
|
# - Zhang, J., and Mitliagkas, I., "YellowFin and the Art of Momentum Tuning," CoRR abs/1706.03471, 2017.
|
12
14
|
class YellowFin
|
@@ -20,6 +22,7 @@ module Rumale
|
|
20
22
|
# @param decay [Float] The smooting parameter.
|
21
23
|
# @param window_width [Integer] The sliding window width for searching curvature range.
|
22
24
|
def initialize(learning_rate: 0.01, momentum: 0.9, decay: 0.999, window_width: 20)
|
25
|
+
warn 'warning: YellowFin is deprecated. This class will be deleted in version 0.20.0.'
|
23
26
|
check_params_numeric(learning_rate: learning_rate, momentum: momentum, decay: decay, window_width: window_width)
|
24
27
|
check_params_positive(learning_rate: learning_rate, momentum: momentum, decay: decay, window_width: window_width)
|
25
28
|
@params = {}
|
@@ -119,6 +119,7 @@ module Rumale
|
|
119
119
|
@steps.keys.reverse_each do |name|
|
120
120
|
transformer = @steps[name]
|
121
121
|
next if transformer.nil?
|
122
|
+
|
122
123
|
itrans_z = transformer.inverse_transform(itrans_z)
|
123
124
|
end
|
124
125
|
itrans_z
|
@@ -140,6 +141,7 @@ module Rumale
|
|
140
141
|
steps.keys[0...-1].each do |name|
|
141
142
|
transformer = steps[name]
|
142
143
|
next if transformer.nil? || %i[fit transform].all? { |m| transformer.class.method_defined?(m) }
|
144
|
+
|
143
145
|
raise TypeError,
|
144
146
|
'Class of intermediate step in pipeline should be implemented fit and transform methods: ' \
|
145
147
|
"#{name} => #{transformer.class}"
|
@@ -158,6 +160,7 @@ module Rumale
|
|
158
160
|
@steps.keys[0...-1].each do |name|
|
159
161
|
transformer = @steps[name]
|
160
162
|
next if transformer.nil?
|
163
|
+
|
161
164
|
transformer.fit(trans_x, y) if fit
|
162
165
|
trans_x = transformer.transform(trans_x)
|
163
166
|
end
|
@@ -5,9 +5,13 @@ require 'rumale/optimizer/nadam'
|
|
5
5
|
|
6
6
|
module Rumale
|
7
7
|
# This module consists of the classes that implement polynomial models.
|
8
|
+
#
|
9
|
+
# @deprecated PolynomialModel module will be deleted in version 0.20.0.
|
8
10
|
module PolynomialModel
|
9
11
|
# BaseFactorizationMachine is an abstract class for implementation of Factorization Machine-based estimators.
|
10
12
|
# This class is used internally.
|
13
|
+
#
|
14
|
+
# @deprecated BaseFactorizationMachine will be deleted in version 0.20.0.
|
11
15
|
class BaseFactorizationMachine
|
12
16
|
include Base::BaseEstimator
|
13
17
|
|
@@ -65,7 +69,7 @@ module Rumale
|
|
65
69
|
factor_optimizers = Array.new(@params[:n_factors]) { @params[:optimizer].dup }
|
66
70
|
# Start optimization.
|
67
71
|
@params[:max_iter].times do |t|
|
68
|
-
sample_ids =
|
72
|
+
sample_ids = Array(0...n_samples)
|
69
73
|
sample_ids.shuffle!(random: sub_rng)
|
70
74
|
until (subset_ids = sample_ids.shift(@params[:batch_size])).empty?
|
71
75
|
# Sampling.
|
@@ -75,6 +79,7 @@ module Rumale
|
|
75
79
|
# Calculate gradients for loss function.
|
76
80
|
loss_grad = loss_gradient(sub_x, ex_sub_x, sub_y, factor_mat, weight_vec)
|
77
81
|
next if loss_grad.ne(0.0).count.zero?
|
82
|
+
|
78
83
|
# Update each parameter.
|
79
84
|
weight_vec = weight_optimizer.call(weight_vec, weight_gradient(loss_grad, ex_sub_x, weight_vec))
|
80
85
|
@params[:n_factors].times do |n|
|