GitHub

DSGU

RecurrentLayers.DSGU — Type
DSGU(input_size => hidden_size;
    return_state = false, kwargs...)

Deep simple gated unit network (Gao and Glowacka, 16–18 Nov 2016). See DSGUCell for a layer that processes a single sequence.

Arguments

  • input_size => hidden_size: input and inner dimension of the layer.

Keyword arguments

  • return_state: Option to return the last state together with the output. Default is false.
  • init_kernel: initializer for the input to hidden weights. Default is glorot_uniform.
  • init_recurrent_kernel: initializer for the hidden to hidden weights. Default is glorot_uniform.
  • bias: include input to recurrent bias or not. Default is true.
  • recurrent_bias: include recurrent to recurrent bias or not. Default is true.
  • independent_recurrence: flag to toggle independent recurrence. If true, the recurrent to recurrent weights are a vector instead of a matrix. Default false.
  • integration_mode: determines how the input and hidden projections are combined. The options are :addition and :multiplicative_integration. Defaults to :addition.

Equations

\[\begin{aligned} \mathbf{x}_g(t) &= \mathbf{W}_{xg} \mathbf{x}(t), \\ \mathbf{z}_g(t) &= \tanh\left(\mathbf{W}_{zg} \left(\mathbf{x}_g(t) \circ \mathbf{h}(t-1)\right) + \mathbf{b}_{zg}\right), \\ \mathbf{z}_{out}(t) &= \sigma\left(\mathbf{W}_{go} \left(\mathbf{z}_g(t) \circ \mathbf{h}(t-1)\right)\right), \\ \mathbf{z}_t &= \operatorname{hard sigmoid}\left( \mathbf{W}_{xz} \mathbf{x}(t) + \mathbf{W}_{hz} \mathbf{h}(t-1) + \mathbf{b}_z\right), \\ \mathbf{h}(t) &= \left(1 - \mathbf{z}_t\right) \circ \mathbf{h}(t-1) + \mathbf{z}_t \circ \mathbf{z}_{out}(t). \end{aligned}\]

Forward

dsgu(inp, state)
dsgu(inp)

Arguments

  • inp: The input to the dsgu. It should be a vector of size input_size x len or a matrix of size input_size x len x batch_size.
  • state: The hidden state of the DSGU. If given, it is a vector of size hidden_size or a matrix of size hidden_size x batch_size. If not provided, it is assumed to be a vector of zeros, initialized by Flux.initialstates.

Returns

  • New hidden states new_states as an array of size hidden_size x len x batch_size. When return_state = true it returns a tuple of the hidden states new_states and the last state of the iteration.
source