Class: Brain::NeuralNetwork

Inherits:
Object
  • Object
show all
Defined in:
lib/brain/neuralnetwork.rb

Instance Method Summary collapse

Constructor Details

#initialize(options = {}) ⇒ NeuralNetwork

Returns a new instance of NeuralNetwork.



8
9
10
11
12
13
 
# File 'lib/brain/neuralnetwork.rb', line 8

def initialize(options = {})
  @learning_rate = options[:learning_rate] || 0.3
  @momentum = options[:momentum] || 0.1
  @hidden_sizes = options[:hidden_layers]
  @binary_thresh = options[:binary_thresh] || 0.5
end

Instance Method Details

#adjust_weights(learning_rate) ⇒ Object 



148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
 
# File 'lib/brain/neuralnetwork.rb', line 148

def adjust_weights(learning_rate)
  (1..@output_layer).each do |layer|
    incoming = @outputs[layer - 1]

    (0...@sizes[layer]).each do |node|
      delta = @deltas[layer][node]

      (0...incoming.length).each do |k|
        change = @changes[layer][node][k]

        change = (learning_rate * delta * incoming[k]) + (@momentum * change)

        @changes[layer][node][k] = change
        @weights[layer][node][k] += change
      end
      @biases[layer][node] += learning_rate * delta
    end
  end
end

#calculate_deltas(target) ⇒ Object 



128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
 
# File 'lib/brain/neuralnetwork.rb', line 128

def calculate_deltas(target)
  (0..@output_layer).to_a.reverse.each do |layer|
    (0...@sizes[layer]).each do |node|
      output = @outputs[layer][node]

      error = 0
      if layer == @output_layer
        error = target[node] - output
      else
        deltas = @deltas[layer + 1]
        (0...deltas.length).each do |k|
          error += deltas[k] * @weights[layer + 1][k][node]
        end
      end
      @errors[layer][node] = error
      @deltas[layer][node] = error * output * (1 - output)
    end
  end
end

#format_data(data) ⇒ Object 



168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
 
# File 'lib/brain/neuralnetwork.rb', line 168

def format_data(data)
  unless data.is_a? Array
    data = [data]
  end

  #turn sparse hash input into arrays with 0s as filler
  unless data[0][:input].is_a? Array
    @input_lookup = Lookup.build_lookup data.map {|d| d[:input]} unless @input_lookup
    data.map! do |datum|
      array = Lookup.to_array @input_lookup, datum[:input]
      datum.merge({ input: array })
    end
  end

  unless data[0][:output].is_a? Array
    @output_lookup = Lookup.build_lookup data.map {|d| d[:output]} unless @output_lookup
    data.map! do |datum|
      array = Lookup.to_array @output_lookup, datum[:output]
      datum.merge({ output: array })
    end
  end

  data
end

#from_json(json) ⇒ Object 



241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
 
# File 'lib/brain/neuralnetwork.rb', line 241

def from_json(json)
  json = JSON.parse(json).deep_symbolize_keys
  size = json[:layers].length


  @output_layer = size - 1
  @sizes = Array.new size
  @weights = Array.new size
  @biases = Array.new size
  @outputs = Array.new size

  (0..@output_layer).each do |i|
    layer = json[:layers][i]
    if i == 0 and (!layer[0] or json[:input_lookup])
      @input_lookup = Lookup.lookup_from_hash layer
    elsif i == @output_layer and (!layer[0] or json[:output_lookup])
      @output_lookup = Lookup.lookup_from_hash layer
    end

    nodes = layer.keys
    @sizes[i] = nodes.length
    @weights[i] = []
    @biases[i] = []
    @outputs[i] = []

    (0...nodes.length).each do |j|

      node = nodes[j]
      @biases[i][j] = layer[node][:bias]
      @weights[i][j] = layer[node][:weights]
      @weights[i][j] = @weights[i][j].values unless @weights[i][j].nil?
    end
  end
end

#init(sizes) ⇒ Object 



15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
 
# File 'lib/brain/neuralnetwork.rb', line 15

def init(sizes)
  @sizes = sizes
  @output_layer = @sizes.length - 1

  @biases = [] # weights for bias nodes
  @weights = []
  @outputs = []

  # state for training
  @deltas = []
  @changes = [] # for momentum
  @errors = []

  (0..@output_layer).each do |layer|
    size = @sizes[layer]
    @deltas[layer] = Array.new size, 0
    @errors[layer] = Array.new size, 0
    @outputs[layer] = Array.new size, 0

    if layer > 0
      @biases[layer] = randos size
      @weights[layer] = Array.new size
      @changes[layer] = Array.new size

      (0...size).each do |node|
        prev_size = @sizes[layer - 1]
        @weights[layer][node] = randos prev_size
        @changes[layer][node] = Array.new prev_size, 0
      end
    end
  end
end

#run(input) ⇒ Object 



48
49
50
51
52
53
54
55
 
# File 'lib/brain/neuralnetwork.rb', line 48

def run(input)
  input = Lookup.to_array(@input_lookup, input) if @input_lookup

  output = run_input input
  output = Lookup.to_hash(@output_lookup, output) if @output_lookup

  output
end

#run_input(input) ⇒ Object 



57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
 
# File 'lib/brain/neuralnetwork.rb', line 57

def run_input(input)
  @outputs[0] = input
  output = 0

  (1..@output_layer).each do |layer|
    (0...@sizes[layer]).each do |node|
      weights = @weights[layer][node]

      sum = @biases[layer][node]
      (0...weights.length).each do |k|
        sum += weights[k] * input[k]
      end
      @outputs[layer][node] = 1 / (1 + Math.exp(-sum))
    end
    output = input = @outputs[layer]
  end

  output
end

#to_json ⇒ Object 



193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
 
# File 'lib/brain/neuralnetwork.rb', line 193

def to_json
  # make json look like:
  # {
  #   layers: [
  #     { x: {},
  #       y: {}},
  #     {'0': {bias: -0.98771313, weights: {x: 0.8374838, y: 1.245858},
  #      '1': {bias: 3.48192004, weights: {x: 1.7825821, y: -2.67899}}},
  #     { f: {bias: 0.27205739, weights: {'0': 1.3161821, '1': 2.00436}}}
  #   ]
  # }
  layers = []
  (0..@output_layer).each do |layer|
    layers[layer] = {}

    if layer == 0 and @input_lookup
      nodes = @input_lookup.keys
    elsif layer == @output_layer and @output_lookup
      nodes = @output_lookup.keys
    else
      nodes = (0...@sizes[layer]).to_a
    end

    (0...nodes.length).each do |j|
      node = nodes[j]
      layers[layer][node] = {}

      if layer > 0
        layers[layer][node][:bias] = @biases[layer][j]
        layers[layer][node][:weights] = {}
        layers[layer - 1].each do |k,v|
          index = k
          if layer == 1 and @input_lookup
            index = @input_lookup[k]
          end
          layers[layer][node][:weights][k] = @weights[layer][j][index]
        end
      end
    end
  end

  {
    layers: layers,
    output_lookup: !!@output_lookup,
    input_lookup: !!@input_lookup
  }.to_json
end

#train(data, options = {}) ⇒ Object 



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
114
115
 
# File 'lib/brain/neuralnetwork.rb', line 77

def train(data, options = {})
  data = format_data data

  iterations = options[:iterations] || 20000
  error_thresh = options[:error_thresh] || 0.003
  log = options[:log] || false
  log_period = options[:log_period] || 10
  learning_rate = options[:learning_rate] || @learning_rate || 0.3

  input_size = data[0][:input].length
  output_size = data[0][:output].length

  hidden_sizes = @hidden_sizes
  hidden_sizes = [[3, (input_size / 2.0).floor].max] unless hidden_sizes
  sizes = [input_size, hidden_sizes, output_size].flatten
  init sizes

  error = 1
  done_iterations = iterations
  (0...iterations).each do |i|
    unless error > error_thresh
      done_iterations = i
      break
    end
    sum = 0
    data.each do |d|
      err = train_pattern d[:input], d[:output], learning_rate
      sum += err
    end
    error = sum / data.length

    puts "iterations: #{i}, training error: #{error}" if log and (i % log_period == 0)
  end

  {
    error: error,
    iterations: done_iterations
  }
end

#train_pattern(input, target, learning_rate) ⇒ Object 



117
118
119
120
121
122
123
124
125
126
 
# File 'lib/brain/neuralnetwork.rb', line 117

def train_pattern(input, target, learning_rate)
  learning_rate ||= @learning_rate

  # forward propogate
  run_input input
  calculate_deltas target
  adjust_weights learning_rate

  mse @errors[@output_layer]
end