spyrit.core.recon.Denoise_layer.tikho

static Denoise_layer.tikho(inputs: tensor, weight: tensor) → tensor

Applies a transformation to the incoming data: \(y = \sigma_\text{prior}^2/(\sigma_\text{prior}^2+x)\).

\(x\) is the input tensor (see inputs) and \(\sigma_\text{prior}\) is the standard deviation prior (see weight).

Args:

inputs (torch.tensor): input tensor \(x\) of shape \((N, *, in\_features)\)

weight (torch.tensor): standard deviation prior \(\sigma_\text{prior}\) of shape \((in\_features)\)

Returns:

torch.tensor: The transformed data \(y\) of shape \((N, in\_features)\)

Shape:

• inputs: \((N, *, in\_features)\) where * means any number of additional dimensions - Variance of measurements

• weight: \((in\_features)\) - corresponds to the standard deviation of our prior.

• output: \((N, in\_features)\)