PoissonLikelihood
- class deepinv.optim.PoissonLikelihood(gain=1.0, bkg=0, normalize=True)[source]
Bases:
DataFidelityPoisson negative log-likelihood.
\[\datafid{z}{y} = -y^{\top} \log(z+\beta)+1^{\top}z\]where \(y\) are the measurements, \(z\) is the estimated (positive) density and \(\beta\geq 0\) is an optional background level.
Note
The function is not Lipschitz smooth w.r.t. \(z\) in the absence of background (\(\beta=0\)).
- Parameters:
bkg (float) – background level \(\beta\).
- d(x, y)[source]
Computes the Poisson negative log-likelihood.
- Parameters:
x (torch.Tensor) – signal \(x\) at which the function is computed.
y (torch.Tensor) – measurement \(y\).
- grad_d(x, y)[source]
Gradient of the Poisson negative log-likelihood.
- Parameters:
x (torch.Tensor) – signal \(x\) at which the function is computed.
y (torch.Tensor) – measurement \(y\).
- prox_d(x, y, gamma=1.0)[source]
Proximal operator of the Poisson negative log-likelihood.
- Parameters:
x (torch.Tensor) – signal \(x\) at which the function is computed.
y (torch.Tensor) – measurement \(y\).
gamma (float) – proximity operator step size.
