View source on GitHub
|
Libraries for constructing federated aggregation.
For example uses of the symbols in this module, see Tuning recommended aggregations for learning tutorial, and for details of the design and how to implement new aggregations, see Implementing Custom Aggregations tutorial.
Classes
class DifferentiallyPrivateFactory: UnweightedAggregationFactory for tensorflow_privacy DPQueries.
class EncodedSumFactory: UnweightedAggregationFactory for encoded sum.
class MeanFactory: Aggregation factory for weighted mean.
class PrivateQuantileEstimationProcess: A tff.templates.EstimationProcess for estimating private quantiles.
class SecureSumFactory: AggregationProcess factory for securely summing values.
class SumFactory: UnweightedAggregationFactory for sum.
class UnweightedAggregationFactory: Factory for creating tff.templates.AggregationProcess without weights.
class UnweightedMeanFactory: Aggregation factory for unweighted mean.
class UnweightedReservoirSamplingFactory: An UnweightedAggregationFactory for reservoir sampling values.
class WeightedAggregationFactory: Factory for creating tff.templates.AggregationProcess with weights.
Functions
clipping_factory(...): Creates an aggregation factory to perform L2 clipping.
federated_max(...): Computes the maximum at tff.SERVER of a value placed at tff.CLIENTS.
federated_min(...): Computes the minimum at tff.SERVER of a value placed at tff.CLIENTS.
federated_sample(...): Aggregation to produce uniform sample of at most max_num_samples values.
secure_quantized_sum(...): Quantizes and sums values securely.
zeroing_factory(...): Creates an aggregation factory to perform zeroing.
Type Aliases
AggregationFactory: The central part of internal API.