Build a trainable LinearOperatorFullMatrix
instance.
tfp.experimental.vi.util.build_trainable_linear_operator_full_matrix(
*args, seed=None, **kwargs
)
Args |
shape
|
Shape of the LinearOperator , equal to [b0, ..., bn, h, w] , where
b0...bn are batch dimensions h and w are the height and width of the
matrix represented by the LinearOperator .
|
scale_initializer
|
Variables are initialized with samples from
Normal(0, scale_initializer) .
|
dtype
|
tf.dtype of the LinearOperator .
|
name
|
str, name for tf.name_scope . seed: PRNG seed; see tfp.random.sanitize_seed for details.
|
Returns |
instance
|
instance parameterized by trainable tf.Variable s.
|