norse.torch.functional.receptive_field.gaussian_kernel

norse.torch.functional.receptive_field.gaussian_kernel(size: int, c: Tensor, x: Tensor, y: Tensor, domain: int = 8) → Tensor

Efficiently creates a differentiable 2d gaussian kernel.

Arguments:

size (int): The size of the kernel c (torch.Tensor): A 2x2 covariance matrix describing the eccentricity of the gaussian x (torch.Tensor): The receptive’s field center position in x-axis y (torch.Tensor): The receptive’s field center position in y-axis domain (int): The domain of the kernel. Defaults to 8 (sampling -8 to 8).