Binary

class colibri.regularizers.Binary(parameter=10, min_v=0, max_v=1)

Bases: Module

Binary Regularization for Coded Aperture Design.

Code adapted from Bacca, Jorge, Tatiana Gelvez-Barrera, and Henry Arguello. “Deep coded aperture design: An end-to-end approach for computational imaging tasks.” IEEE Transactions on Computational Imaging 7 (2021): 1148-1160. The regularizer computes:

\[\begin{equation*} R(\learnedOptics) = \mu\sum_{i=1}^{n} (\learnedOptics_i - \text{min_v})^2(\learnedOptics_i - \text{max_v})^2 \end{equation*}\]

where \(\mu\) is the regularization parameter and \(\text{min_v}\) and \(\text{max_v}\) are the minimum and maximum values for the weights, respectively.

Parameters:

• parameter (float) – Regularization parameter.

• min_v (float) – Minimum value for weight clipping.

• max_v (float) – Maximum value for weight clipping.

forward(x)

Compute binary regularization term.

Parameters:

x (torch.Tensor) – Input tensor (layer’s weight).

Returns:

Binary regularization term.

Return type:

torch.Tensor

Examples using Binary:

Demo Colibri.

Demo Colibri.