Returns a kernel Keras initializer object from its id.
tfl.lattice_layer.create_kernel_initializer(
kernel_initializer_id,
lattice_sizes,
monotonicities,
output_min,
output_max,
unimodalities,
joint_unimodalities,
init_min=None,
init_max=None
)
This function is used to convert the 'kernel_initializer' parameter in the
constructor of tfl.Lattice into the corresponding initializer object.
Args |
kernel_initializer_id
|
See the documentation of the 'kernel_initializer'
parameter in the constructor of tfl.Lattice.
|
lattice_sizes
|
See the documentation of the same parameter in the
constructor of tfl.Lattice.
|
monotonicities
|
See the documentation of the same parameter in the
constructor of tfl.Lattice.
|
output_min
|
See the documentation of the same parameter in the constructor
of tfl.Lattice.
|
output_max
|
See the documentation of the same parameter in the constructor
of tfl.Lattice.
|
unimodalities
|
See the documentation of the same parameter in the
constructor of tfl.Lattice.
|
joint_unimodalities
|
See the documentation of the same parameter in the
constructor of tfl.Lattice.
|
init_min
|
None or lower bound of kernel initialization. If set, init_max
must also be set.
|
init_max
|
None or upper bound of kernel initialization. If set, init_min
must also be set.
|
Returns |
The Keras initializer object for the tfl.Lattice kernel variable.
|
Raises |
ValueError
|
If only one of init_{min/max} is set.
|