Class: DNN::Layers::Dense

Inherits:

Connection

• Object
• Layer
• HasParamLayer
• Connection
• DNN::Layers::Dense

Defined in:

lib/dnn/core/layers.rb

Overview

Full connnection layer.

Instance Attribute Summary

• #num_nodes ⇒ Integer readonly

  Number of nodes.

Attributes inherited from Connection

#bias_initializer, #bias_regularizer, #weight_initializer, #weight_regularizer

Attributes inherited from HasParamLayer

#params, #trainable

Attributes inherited from Layer

#input_shape, #learning_phase

Class Method Summary

• .from_hash(hash) ⇒ Object

Instance Method Summary

• #backward(dy) ⇒ Object
• #build(input_shape) ⇒ Object
• #forward(x) ⇒ Object
• #initialize(num_nodes, weight_initializer: Initializers::RandomNormal.new, bias_initializer: Initializers::Zeros.new, weight_regularizer: nil, bias_regularizer: nil, use_bias: true) ⇒ Dense constructor

  A new instance of Dense.

• #output_shape ⇒ Object
• #to_hash ⇒ Object

Methods inherited from Connection

#regularizers, #use_bias

Methods inherited from Layer

#built?

Constructor Details

#initialize(num_nodes, weight_initializer: Initializers::RandomNormal.new, bias_initializer: Initializers::Zeros.new, weight_regularizer: nil, bias_regularizer: nil, use_bias: true) ⇒ Dense

Returns a new instance of Dense.

Parameters:

• num_nodes (Integer) —

  number of nodes.

# File 'lib/dnn/core/layers.rb', line 179

def initialize(num_nodes,
               weight_initializer: Initializers::RandomNormal.new,
               bias_initializer: Initializers::Zeros.new,
               weight_regularizer: nil,
               bias_regularizer: nil,
               use_bias: true)
  super(weight_initializer: weight_initializer, bias_initializer: bias_initializer,
        weight_regularizer: weight_regularizer, bias_regularizer: bias_regularizer, use_bias: use_bias)
  @num_nodes = num_nodes
end

Instance Attribute Details

#num_nodes ⇒ Integer (readonly)

Returns number of nodes.

Returns:

• (Integer) —

  number of nodes.

# File 'lib/dnn/core/layers.rb', line 167

def num_nodes
  @num_nodes
end

Class Method Details

.from_hash(hash) ⇒ Object

# File 'lib/dnn/core/layers.rb', line 169

def self.from_hash(hash)
  self.new(hash[:num_nodes],
           weight_initializer: Utils.from_hash(hash[:weight_initializer]),
           bias_initializer: Utils.from_hash(hash[:bias_initializer]),
           weight_regularizer: Utils.from_hash(hash[:weight_regularizer]),
           bias_regularizer: Utils.from_hash(hash[:bias_regularizer]),
           use_bias: hash[:use_bias])
end

Instance Method Details

#backward(dy) ⇒ Object

# File 'lib/dnn/core/layers.rb', line 205

def backward(dy)
  if @trainable
    @weight.grad += @x.transpose.dot(dy)
    @bias.grad += dy.sum(0) if @bias
  end
  dy.dot(@weight.data.transpose)
end

#build(input_shape) ⇒ Object

# File 'lib/dnn/core/layers.rb', line 190

def build(input_shape)
  super
  num_prev_nodes = input_shape[0]
  @weight.data = Xumo::SFloat.new(num_prev_nodes, @num_nodes)
  @bias.data = Xumo::SFloat.new(@num_nodes) if @bias
  init_weight_and_bias
end

#forward(x) ⇒ Object

# File 'lib/dnn/core/layers.rb', line 198

def forward(x)
  @x = x
  y = x.dot(@weight.data)
  y += @bias.data if @bias
  y
end

#output_shape ⇒ Object

# File 'lib/dnn/core/layers.rb', line 213

def output_shape
  [@num_nodes]
end

#to_hash ⇒ Object

# File 'lib/dnn/core/layers.rb', line 217

def to_hash
  super({num_nodes: @num_nodes})
end