Save the date! Google I/O returns May 18-20 Register now

tfl.lattice_lib.verify_hyperparameters

Verifies that all given hyperparameters are consistent.

tfl.lattice_lib.verify_hyperparameters(
    lattice_sizes, units=None, weights_shape=None, input_shape=None,
    monotonicities=None, unimodalities=None, edgeworth_trusts=None,
    trapezoid_trusts=None, monotonic_dominances=None, range_dominances=None,
    joint_monotonicities=None, joint_unimodalities=None, output_min=None,
    output_max=None, regularization_amount=None, regularization_info='',
    interpolation='hypercube'
)

This function does not inspect weights themselves. Only their shape. Use assert_constraints() to assert actual weights against constraints.

See tfl.layers.Lattice class level comment for detailed description of arguments.

lattice_sizes Lattice sizes to check againts.
units Units hyperparameter of Lattice layer.
weights_shape Shape of tensor which represents Lattice layer weights.
input_shape Shape of layer input. Useful only if units is set.
monotonicities Monotonicities hyperparameter of Lattice layer.
unimodalities Unimodalities hyperparameter of Lattice layer.
edgeworth_trusts Edgeworth_trusts hyperparameter of Lattice layer.
trapezoid_trusts Trapezoid_trusts hyperparameter of Lattice layer.
monotonic_dominances Monotonic dominances hyperparameter of Lattice layer.
range_dominances Range dominances hyperparameter of Lattice layer.
joint_monotonicities Joint monotonicities hyperparameter of Lattice layer.
joint_unimodalities Joint unimodalities hyperparameter of Lattice layer.
output_min Minimum output of Lattice layer.
output_max Maximum output of Lattice layer.
regularization_amount Regularization amount for regularizers.
regularization_info String which describes regularization_amount.
interpolation One of 'simplex' or 'hypercube' interpolation.

ValueError If something is inconsistent.