|View source on GitHub|
Implements CDF transformation with derived parameters (kernels).
cdf_fn is similar to
tfl.layers.CDF, which is an additive / multiplicative
average of a few shifted and scaled
relu6 basis functions,
with the difference that the functions are parametrized by the provided
parameters instead of learnable weights belonging to a
These parameters can be one of:
- trainable variables,
- outputs from other TF modules.
For inputs of shape
(batch_size, input_dim), two sets of free-form
parameters are used to configure the CDF function:
location_parametersfor where to place the sigmoid / relu6 transformation basis,
scaling_parameters(optional) for the horizontal scaling before applying the transformation basis.
inputs through a CDF function specified by keypoint parameters.