Module: tfl.lattice_layer

Lattice layer with monotonicity, unimodality, trust and bound constraints.

Keras implementation of tensorflow lattice layer. This layer takes one or more d-dimensional input(s) and combines them using a lattice function, satisfying monotonicity, unimodality, trust and bound constraints if specified.

Classes

class LaplacianRegularizer: Laplacian regularizer for tfl.layers.Lattice layer.

class LatticeConstraints: Constraints for tfl.layers.Lattice layer.

class LinearInitializer: Initializes a tfl.layers.Lattice as linear function.

class RandomMonotonicInitializer: Initializes a tfl.layers.Lattice as uniform random monotonic function.

class TorsionRegularizer: Torsion regularizer for tfl.layers.Lattice layer.

Functions

create_kernel_initializer(...): Returns a kernel Keras initializer object from its id.

Other Members

  • LATTICE_KERNEL_NAME = 'lattice_kernel'
  • LATTICE_SIZES_NAME = 'lattice_sizes'