p2pfl.learning.aggregators.scaffold module

Callback for SCAFFOLD operations.

class p2pfl.learning.aggregators.scaffold.Scaffold(global_lr=0.1, disable_partial_aggregation=False)

Bases: Aggregator

SCAFFOLD Aggregator.

Paper: https://arxiv.org/pdf/1910.06378 The aggregator acts like the server in centralized learning, handling both model and control variate updates.

Due to the complete decentralization of the enviroment, a global model is also maintained in the aggregator. This consumes additional bandwidth.

::todo:: Improve efficiency by sharing the global model only each n rounds.

Parameters:

• global_lr (float)

• disable_partial_aggregation (bool)

REQUIRED_INFO_KEYS = ['delta_y_i', 'delta_c_i']

SUPPORTS_PARTIAL_AGGREGATION: bool = False

aggregate(models)

Aggregate the models and control variates from clients.

Parameters:

models (list[P2PFLModel]) – List of models to aggregate.

Return type:

P2PFLModel

get_required_callbacks()

Retrieve the list of required callback keys for this aggregator.

Return type:

list[str]