Class: Rumale::LinearModel::SVC

Inherits:
BaseLinearModel show all
Includes:
Base::Classifier
Defined in:
lib/rumale/linear_model/svc.rb

Overview

SVC is a class that implements Support Vector Classifier with mini-batch stochastic gradient descent optimization. For multiclass classification problem, it uses one-vs-the-rest strategy.

Reference

    1. Shalev-Shwartz and Y. Singer, “Pegasos: Primal Estimated sub-GrAdient SOlver for SVM,” Proc. ICML’07, pp. 807–814, 2007.

Examples:

estimator =
  Rumale::LinearModel::SVC.new(reg_param: 1.0, max_iter: 1000, batch_size: 20, random_seed: 1)
estimator.fit(training_samples, traininig_labels)
results = estimator.predict(testing_samples)

Instance Attribute Summary collapse

Attributes included from Base::BaseEstimator

#params

Instance Method Summary collapse

Methods included from Base::Classifier

#score

Constructor Details

#initialize(reg_param: 1.0, fit_bias: false, bias_scale: 1.0, max_iter: 1000, batch_size: 20, probability: false, optimizer: nil, n_jobs: nil, random_seed: nil) ⇒ SVC

Create a new classifier with Support Vector Machine by the SGD optimization.



56
57
58
59
60
61
62
63
64
65
66
67
68
69
 
# File 'lib/rumale/linear_model/svc.rb', line 56

def initialize(reg_param: 1.0, fit_bias: false, bias_scale: 1.0,
               max_iter: 1000, batch_size: 20, probability: false, optimizer: nil, n_jobs: nil, random_seed: nil)
  check_params_float(reg_param: reg_param, bias_scale: bias_scale)
  check_params_integer(max_iter: max_iter, batch_size: batch_size)
  check_params_boolean(fit_bias: fit_bias, probability: probability)
  check_params_type_or_nil(Integer, n_jobs: n_jobs, random_seed: random_seed)
  check_params_positive(reg_param: reg_param, bias_scale: bias_scale, max_iter: max_iter, batch_size: batch_size)
  keywd_args = method(:initialize).parameters.map { |_t, arg| [arg, binding.local_variable_get(arg)] }.to_h
  keywd_args.delete(:probability)
  super(keywd_args)
  @params[:probability] = probability
  @prob_param = nil
  @classes = nil
end

Instance Attribute Details

#bias_term ⇒ Numo::DFloat (readonly)

Return the bias term (a.k.a. intercept) for SVC.



31
32
33
 
# File 'lib/rumale/linear_model/svc.rb', line 31

def bias_term
  @bias_term
end

#classes ⇒ Numo::Int32 (readonly)

Return the class labels.



35
36
37
 
# File 'lib/rumale/linear_model/svc.rb', line 35

def classes
  @classes
end

#rng ⇒ Random (readonly)

Return the random generator for performing random sampling.



39
40
41
 
# File 'lib/rumale/linear_model/svc.rb', line 39

def rng
  @rng
end

#weight_vec ⇒ Numo::DFloat (readonly)

Return the weight vector for SVC.



27
28
29
 
# File 'lib/rumale/linear_model/svc.rb', line 27

def weight_vec
  @weight_vec
end

Instance Method Details

#decision_function(x) ⇒ Numo::DFloat

Calculate confidence scores for samples.



119
120
121
122
 
# File 'lib/rumale/linear_model/svc.rb', line 119

def decision_function(x)
  check_sample_array(x)
  x.dot(@weight_vec.transpose) + @bias_term
end

#fit(x, y) ⇒ SVC

Fit the model with given training data.



76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
 
# File 'lib/rumale/linear_model/svc.rb', line 76

def fit(x, y)
  check_sample_array(x)
  check_label_array(y)
  check_sample_label_size(x, y)

  @classes = Numo::Int32[*y.to_a.uniq.sort]
  n_classes = @classes.size
  n_features = x.shape[1]

  if n_classes > 2
    # initialize model.
    @weight_vec = Numo::DFloat.zeros(n_classes, n_features)
    @bias_term = Numo::DFloat.zeros(n_classes)
    @prob_param = Numo::DFloat.zeros(n_classes, 2)
    # fit model.
    models = if enable_parallel?
               # :nocov:
               parallel_map(n_classes) do |n|
                 bin_y = Numo::Int32.cast(y.eq(@classes[n])) * 2 - 1
                 partial_fit(x, bin_y)
               end
               # :nocov:
             else
               Array.new(n_classes) do |n|
                 bin_y = Numo::Int32.cast(y.eq(@classes[n])) * 2 - 1
                 partial_fit(x, bin_y)
               end
             end
    # store model.
    models.each_with_index { |model, n| @weight_vec[n, true], @bias_term[n], @prob_param[n, true] = model }
  else
    negative_label = y.to_a.uniq.min
    bin_y = Numo::Int32.cast(y.ne(negative_label)) * 2 - 1
    @weight_vec, @bias_term, @prob_param = partial_fit(x, bin_y)
  end

  self
end

#marshal_dump ⇒ Hash

Dump marshal data.



164
165
166
167
168
169
170
171
 
# File 'lib/rumale/linear_model/svc.rb', line 164

def marshal_dump
  { params: @params,
    weight_vec: @weight_vec,
    bias_term: @bias_term,
    prob_param: @prob_param,
    classes: @classes,
    rng: @rng }
end

#marshal_load(obj) ⇒ nil

Load marshal data.



175
176
177
178
179
180
181
182
183
 
# File 'lib/rumale/linear_model/svc.rb', line 175

def marshal_load(obj)
  @params = obj[:params]
  @weight_vec = obj[:weight_vec]
  @bias_term = obj[:bias_term]
  @prob_param = obj[:prob_param]
  @classes = obj[:classes]
  @rng = obj[:rng]
  nil
end

#predict(x) ⇒ Numo::Int32

Predict class labels for samples.



128
129
130
131
132
133
134
135
136
137
138
139
140
141
 
# File 'lib/rumale/linear_model/svc.rb', line 128

def predict(x)
  check_sample_array(x)

  return Numo::Int32.cast(decision_function(x).ge(0.0)) * 2 - 1 if @classes.size <= 2

  n_samples, = x.shape
  decision_values = decision_function(x)
  predicted = if enable_parallel?
                parallel_map(n_samples) { |n| @classes[decision_values[n, true].max_index] }
              else
                Array.new(n_samples) { |n| @classes[decision_values[n, true].max_index] }
              end
  Numo::Int32.asarray(predicted)
end

#predict_proba(x) ⇒ Numo::DFloat

Predict probability for samples.



147
148
149
150
151
152
153
154
155
156
157
158
159
160
 
# File 'lib/rumale/linear_model/svc.rb', line 147

def predict_proba(x)
  check_sample_array(x)

  if @classes.size > 2
    probs = 1.0 / (Numo::NMath.exp(@prob_param[true, 0] * decision_function(x) + @prob_param[true, 1]) + 1.0)
    return (probs.transpose / probs.sum(axis: 1)).transpose
  end

  n_samples, = x.shape
  probs = Numo::DFloat.zeros(n_samples, 2)
  probs[true, 1] = 1.0 / (Numo::NMath.exp(@prob_param[0] * decision_function(x) + @prob_param[1]) + 1.0)
  probs[true, 0] = 1.0 - probs[true, 1]
  probs
end