RecurrentLayer

layers.RecurrentLayer.RecurrentLayer(
    self
    size
    tau=10.0
    transfer_function='tanh'
)

Reservoir of recurrently connected neurons.

\tau \, \frac{d \mathbf{x}(t)}{dt} + \mathbf{x}(t) = W^\text{in} \times I(t) + W^\text{rec} \times \mathbf{r}(t) + W^\text{fb} \times \mathbf{z}(t)

\mathbf{r}(t) = f(\mathbf{x}(t))

Parameters

Name	Type	Description	Default
size	int	number of neurons.	required
tau	float	time constant.	10.0
transfer_function	str	transfer function.	'tanh'

Methods

Name	Description
output
reset	Resets the vectors x and r to 0.
step	Performs one update of the internal variables.

output

layers.RecurrentLayer.RecurrentLayer.output()

Returns

Name	Type	Description
	None	a vector of activities.

reset

layers.RecurrentLayer.RecurrentLayer.reset()

Resets the vectors x and r to 0.

step

layers.RecurrentLayer.RecurrentLayer.step()

Performs one update of the internal variables.