NBRCell

NBRCell(in_dims => out_dims;
    use_bias=true, use_recurrent_ bias=true, train_state=false, init_bias=nothing,
    init_weight=nothing, init_recurrent_weight=nothing,
    init_state=zeros32)

Recurrently neuromodulated bistable recurrent cell.

Equations

\[\begin{aligned} \mathbf{a}(t) &= 1 + \tanh\left(\mathbf{W}_{ih}^{a} \mathbf{x}(t) + \mathbf{b}_{ih}^a + \mathbf{W}_{hh}^{a} \circ \mathbf{h}(t-1)+ \mathbf{b}_{hh}^a \right) \\ \mathbf{c}(t) &= \sigma\left(\mathbf{W}_{ih}^{c} \mathbf{x}(t) + \mathbf{b}_{ih}^c + \mathbf{W}_{hh}^{c} \circ \mathbf{h}(t-1) + \mathbf{b}_{hh}^c \right)\\ \mathbf{h}(t) &= \mathbf{c}(t) \circ \mathbf{h}(t-1) + (1 - \mathbf{c}(t)) \circ \tanh\left(\mathbf{W}_{ih}^{h} \mathbf{x}(t) + \mathbf{b}_{ih}^h + \mathbf{a}(t) \circ \mathbf{h}(t-1)\right) \end{aligned}\]

Arguments

• in_dims: Input Dimension
• out_dims: Output (Hidden State & Memory) Dimension

Keyword Arguments

• use_bias: Flag to use bias $\mathbf{b}_{ih}$ in the computation. Default set to true.
• use_recurrent_bias: Flag to use recurrent bias $\mathbf{b}_{hh}$ in the computation. Default set to true.
• train_state: Flag to set the initial hidden state as trainable. Default set to false.
• init_bias: Initializer for input to hidden bias $\mathbf{b}_{ih}^a, \mathbf{b}_{ih}^c, \mathbf{b}_{ih}^h$. Must be a tuple containing 3 functions, e.g., (glorot_normal, kaiming_uniform). If a single function fn is provided, it is automatically expanded into a 3-element tuple (fn, fn). If set to nothing, weights are initialized from a uniform distribution within [-bound, bound] where bound = inv(sqrt(out_dims)). Default is nothing.
• init_recurrent_bias: Initializer for hidden to hidden bias $\mathbf{b}_{hh}^a, \mathbf{b}_{hh}^c$. Must be a tuple containing 2 functions, e.g., (glorot_normal, kaiming_uniform). If a single function fn is provided, it is automatically expanded into a 2-element tuple (fn, fn). If set to nothing, weights are initialized from a uniform distribution within [-bound, bound] where bound = inv(sqrt(out_dims)). Default is nothing.
• init_weight: Initializer for input to hidden weights $\mathbf{W}_{ih}^a, \mathbf{W}_{ih}^c, \mathbf{W}_{ih}^h$. Must be a tuple containing 3 functions, e.g., (glorot_normal, kaiming_uniform). If a single function fn is provided, it is automatically expanded into a 3-element tuple (fn, fn). If set to nothing, weights are initialized from a uniform distribution within [-bound, bound] where bound = inv(sqrt(out_dims)). Default is nothing.
• init_recurrent_weight: Initializer for input to hidden weights $\mathbf{W}_{hh}^a, \mathbf{W}_{hh}^c$. Must be a tuple containing 2 functions, e.g., (glorot_normal, kaiming_uniform). If a single function fn is provided, it is automatically expanded into a 2-element tuple (fn, fn). If set to nothing, weights are initialized from a uniform distribution within [-bound, bound] where bound = inv(sqrt(out_dims)). Default is nothing.
• init_state: Initializer for hidden state. Default set to zeros32.

Inputs

• Case 1a: Only a single input x of shape (in_dims, batch_size), train_state is set to false - Creates a hidden state using init_state and proceeds to Case 2.
• Case 1b: Only a single input x of shape (in_dims, batch_size), train_state is set to true - Repeats hidden_state from parameters to match the shape of x and proceeds to Case 2.
• Case 2: Tuple (x, (h, )) is provided, then the output and a tuple containing the updated hidden state is returned.

Returns

• Tuple containing

  • Output $h_{new}$ of shape (out_dims, batch_size)
  • Tuple containing new hidden state $h_{new}$

• Updated model state

Parameters

• weight_ih: Concatenated weights to map from input to the hidden state $\{ \mathbf{W}_{ih}^a, \mathbf{W}_{ih}^c, \mathbf{W}_{ih}^h \}$ The initializers in init_weight are applied in the order they appear: the first function is used for $\mathbf{W}_{ih}^a$, the second for $\mathbf{W}_{ih}^c$, and the third for $\mathbf{W}_{ih}^h$.
• weight_hh: Weights to map the hidden state to the hidden state $\{ \mathbf{W}_{hh}^a, \mathbf{W}_{hh}^c \}$ The initializers in init_weight are applied in the order they appear: the first function is used for $\mathbf{W}_{hh}^a$, and the second for $\mathbf{W}_{hh}^c$.
• bias_ih: Bias vector for the input-hidden connection (not present if use_bias=false) $\{ \mathbf{b}_{ih}^a, \mathbf{b}_{ih}^c, \mathbf{b}_{ih}^h \}$ The initializers in init_bias are applied in the order they appear: the first function is used for $\mathbf{b}_{ih}^z$, the second for $\mathbf{b}_{ih}^c$, and the third for $\mathbf{b}_{ih}^h$.
• bias_hh: Bias vector for the input-hidden connection (not present if use_bias=false) $\{ \mathbf{b}_{hh}^a, \mathbf{b}_{hh}^c \}$ The initializers in init_bias are applied in the order they appear: the first function is used for $\mathbf{b}_{hh}^z$, and the second for $\mathbf{b}_{hh}^c$.
• hidden_state: Initial hidden state vector (not present if train_state=false)

States

• rng: Controls the randomness (if any) in the initial state generation