View source on GitHub

Config for aggregate function learning model.

Used in the notebooks

Used in the tutorials

An aggregate function learning model applies piecewise-linear and categorical calibration on the ragged input features, followed by an aggregation layer that aggregates the calibrated inputs. Lastly a lattice model and an optional output piecewise-linear calibration are applied.


model_config = tfl.configs.AggregateFunctionConfig(
model = tfl.premade.AggregateFunction(model_config)

feature_configs A list of tfl.configs.FeatureConfig instances that specify configurations for each feature.
regularizer_configs A list of tfl.configs.RegularizerConfig instances that apply global regularization.
middle_dimension The number of calibrated lattices that are applied to each block. The outputs of these lattices are then averaged over the blocks, and the middle_dimension resulting numbers are then passed into the "middle" calibrated lattice. This middle lattice therefore has input dimension equal to middle_dimension.
middle_lattice_size Size of each of the middle_lattice dimensions.
middle_calibration If a piecewise-linear calibration should be used on the inputs to the middle lattice.
middle_calibration_num_keypoints Number of keypoints to use for the middle piecewise-linear calibration.
middle_monotonicity Specifies if the middle calibrators should be monotonic, using 'increasing' or 1 to indicate increasing monotonicity, 'decreasing' or -1 to indicate decreasing monotonicity, and 'none' or 0 to indicate no monotonicity constraints.
output_min Lower bound constraint on the output of the model.
output_max Upper bound constraint on the output of the model.
output_calibration If a piecewise-linear calibration should be used on the output of the lattice.
output_calibration_num_keypoints Number of keypoints to use for the output piecewise-linear calibration.
output_initialization The initial values to setup for the output of the model. When using output calibration, these values are used to initialize the output keypoints of the output piecewise-linear calibration. Otherwise the lattice parameters will be setup to form a linear function in the range of output_initialization. It can be one of:

  • String 'uniform': Output is initliazed uniformly in label range.
  • A list of numbers: To be used for initialization of the output lattice or output calibrator.



View source

Returns a deserialized configuration dictionary.


View source

Returns existing or default FeatureConfig with the given name.


View source


View source

Returns a configuration dictionary.


View source

Returns existing or default RegularizerConfig with the given name.