ring#

torchdyno.models.initializers.ring(n_units, values=None, dtype=torch.float32)[source]#
Ring matrix. See:

C. Gallicchio & A. Micheli (2020). Ring Reservoir Neural Networks for Graphs. In 2020 International Joint Conference on Neural Networks (IJCNN), IEEE. https://doi.org/10.1109/IJCNN48605.2020.9206723

Parameters:

  • shape (torch.Size) – shape of the tensor.

  • rho (Optional[float], optional) – spectral radius value. Defaults to None.

  • sigma (Optional[float], optional) – standard deviation used for normal initialization. Defaults to None.

  • scale (Optional[float], optional) – scaling value. Defaults to None.

  • n_units (int)

  • values (float | Tensor | None)

  • dtype (dtype)

Returns:

initialized ring tensor.

Return type:

torch.Tensor