spyrit.core.prep.UnsplitRescaleEstim.sigma

UnsplitRescaleEstim.sigma(y: tensor) → tensor

Estimate the variance of raw split measurements as

\[\sigma^2 = \frac{y_+ + y_-}{\alpha^2}\]

where \(y_-\) \(y_+\) are the raw measurements.

Important

This function takes the raw measurments as input and must be called before forward().

Note

alpha could be saved to avoid to recomputing it.

Args:

y (torch.tensor): batch of measurements of shape \((*, 2M)\).

Returns:

torch.tensor: Estimated variance with shape \((*, M)\).