Thanks for tuning in to Google I/O. View all sessions on demandWatch on demand

tfl.kronecker_factored_lattice_lib.scale_initializer

Initializes scale depending on output_min and output_max.

tfl.kronecker_factored_lattice_lib.scale_initializer(
    units, num_terms, output_min, output_max
)

If both output_min and output_max are set, scale is initialized to half their difference, alternating signs for each term. If only output_min is set, scale is initialized to 1 for each term. If only output_max is set, scale is initialized to -1 for each term. Otherwise scale is initialized to alternate between 1 and -1 for each term.

units Output dimension of the layer. Each unit's scale will be initialized identically.
num_terms Number of independently trained submodels per unit, the outputs of which are averaged to get the final output.
output_min None or minimum layer output.
output_max None or maximum layer output.

Kronecker-Factored Lattice scale of shape: (units, num_terms).