ZINB
latent.losses.ZINB
Computes zero-inflated negative binomial loss between y_true and y_pred given
a dispersion parameter (theta) and dropout rate (pi).
__init__(self, pi, theta, eps=1e-08, **kwargs)
special
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
theta |
Union[tensorflow.python.framework.ops.Tensor, float] |
Positive float. Dispersion parameter. |
required |
pi |
Union[tensorflow.python.framework.ops.Tensor, float] |
Positive float between 0 and 1. Dropout rate. |
required |
eps |
float |
Positive float. Clipping value for numerical stability. |
1e-08 |
**kwargs |
|
Other arguments passed to |
{} |