gan_dgmr

Module for various loss functions used with DGMR GAN.

class mfai.pytorch.losses.gan_dgmr.GridCellLoss(precip_weight_cap=24.0)

Bases: Module

Grid Cell Regularizer loss from Skillful Nowcasting, see equation (3) of the original paper (https://arxiv.org/pdf/2104.00954.pdf).

Parameters:

precip_weight_cap (float)

forward(generated_images, targets)

Forward function.

Calculates the grid cell regularizer value, assumes generated images are the mean predictions from 6 calls to the generator (Monte Carlo estimation of the expectations for the latent variable).

\[\begin{split}L_R(\\Theta) = \frac{1}{HWN} \\| (\\mathbb{E}_Z [G_|theta(Z; X_{1:M})] - X_{M+1:M+N}) \\circ w(X_{M+1:M+N}) \\|_1\end{split}\]

where H, W and N represent height, width and leadtimes.

Parameters:

• generated_images (Tensor) – Mean generated images from the generator

• targets (Tensor) – Ground truth future frames

Return type:

Tensor

Returns:

Grid Cell Regularizer term

weight_fn(y, precip_weight_cap=24.0)

Weight function for the grid cell loss.

w(y) = max(y + 1, precip_weight_cap)

Parameters:

• y (Tensor) – Tensor of rainfall intensities.

• precip_weight_cap (float) – Custom ceiling for the weight function.

Return type:

Tensor

Returns:

Weights for each grid cell.

mfai.pytorch.losses.gan_dgmr.loss_hinge_disc(score_generated, score_real)

Discriminator Hinge loss.

Return type:

Tensor

Parameters:

• score_generated (Tensor)

• score_real (Tensor)

mfai.pytorch.losses.gan_dgmr.loss_hinge_gen(score_generated)

Generator Hinge loss.

Return type:

Tensor

Parameters:

score_generated (Tensor)