openfl.interface.aggregation_functions

class openfl.interface.aggregation_functions.AdagradAdaptiveAggregation(*args, **kwargs)

Adagrad adaptive Federated Aggregation funtcion.

class openfl.interface.aggregation_functions.AdamAdaptiveAggregation(*args, **kwargs)

Adam adaptive Federated Aggregation funtcion.

class openfl.interface.aggregation_functions.AggregationFunction(*args, **kwargs)

Interface for specifying aggregation function.

abstract call(local_tensors: List[LocalTensor], db_iterator: Iterator[Series], tensor_name: str, fl_round: int, tags: Tuple[str]) → ndarray

Aggregate tensors.

Parameters:

• local_tensors (list[openfl.utilities.LocalTensor]) – List of local tensors to aggregate.

• db_iterator –

  An iterator over history of all tensors. Columns: - ‘tensor_name’: name of the tensor.

  Examples for `torch.nn.Module`s: ‘conv1.weight’,’fc2.bias’.

  • ’round’: 0-based number of round corresponding to this

    tensor.

  • ’tags’: tuple of tensor tags. Tags that can appear:

    • ’model’ indicates that the tensor is a model parameter.

    • ’trained’ indicates that tensor is a part of a training

      result. These tensors are passed to the aggregator node after local learning.

    • ’aggregated’ indicates that tensor is a result of

      aggregation. These tensors are sent to collaborators for the next round.

    • ’delta’ indicates that value is a difference between

      rounds for a specific tensor.

    also one of the tags is a collaborator name if it corresponds to a result of a local task.

  • ’nparray’: value of the tensor.

• tensor_name – name of the tensor

• fl_round – round number

• tags – tuple of tags for this tensor

Returns:

np.ndarray – aggregated tensor

class openfl.interface.aggregation_functions.FedCurvWeightedAverage(*args, **kwargs)

Aggregation function of FedCurv algorithm.

Applies weighted average aggregation to all tensors except Fisher matrices variables (u_t, v_t). These variables are summed without weights.

FedCurv paper: https://arxiv.org/pdf/1910.07796.pdf

call(local_tensors, tensor_db, tensor_name, fl_round, tags)

Apply aggregation.

class openfl.interface.aggregation_functions.GeometricMedian(*args, **kwargs)

Geometric median aggregation.

call(local_tensors, *_) → ndarray

Aggregate tensors.

Parameters:

• local_tensors (list[openfl.utilities.LocalTensor]) – List of local tensors to aggregate.

• db_iterator –

  iterator over history of all tensors. Columns: - ‘tensor_name’: name of the tensor.

  Examples for `torch.nn.Module`s: ‘conv1.weight’, ‘fc2.bias’.

  • ’round’: 0-based number of round corresponding to this

    tensor.

  • ’tags’: tuple of tensor tags. Tags that can appear:

    • ’model’ indicates that the tensor is a model parameter.

    • ’trained’ indicates that tensor is a part of a training

      result. These tensors are passed to the aggregator node after local learning.

    • ’aggregated’ indicates that tensor is a result of

      aggregation. These tensors are sent to collaborators for the next round.

    • ’delta’ indicates that value is a difference between

      rounds for a specific tensor.

    also one of the tags is a collaborator name if it corresponds to a result of a local task.

  • ’nparray’: value of the tensor.

• tensor_name – name of the tensor

• fl_round – round number

• tags – tuple of tags for this tensor

Returns:

geometric_median (np.ndarray) – aggregated tensor

class openfl.interface.aggregation_functions.Median(*args, **kwargs)

Median aggregation.

call(local_tensors, *_) → ndarray

Aggregate tensors.

Parameters:

• local_tensors (list[openfl.utilities.LocalTensor]) – List of local tensors to aggregate.

• db_iterator –

  iterator over history of all tensors. Columns: - ‘tensor_name’: name of the tensor.

  Examples for `torch.nn.Module`s: ‘conv1.weight’,’fc2.bias’.

  • ’round’: 0-based number of round corresponding to this

    tensor.

  • ’tags’: tuple of tensor tags. Tags that can appear:

    • ’model’ indicates that the tensor is a model parameter.

    • ’trained’ indicates that tensor is a part of a training

      result. These tensors are passed to the aggregator node after local learning.

    • ’aggregated’ indicates that tensor is a result of

      aggregation. These tensors are sent to collaborators for the next round.

    • ’delta’ indicates that value is a difference between

      rounds for a specific tensor.

    also one of the tags is a collaborator name if it corresponds to a result of a local task.

  • ’nparray’: value of the tensor.

• tensor_name – name of the tensor

• fl_round – round number

• tags – tuple of tags for this tensor

Returns:

np.ndarray – aggregated tensor

class openfl.interface.aggregation_functions.WeightedAverage(*args, **kwargs)

Weighted average aggregation.

call(local_tensors, *_) → ndarray

Aggregate tensors.

Parameters:

• local_tensors (list[openfl.utilities.LocalTensor]) – List of local tensors to aggregate.

• db_iterator –

  iterator over history of all tensors. Columns: - ‘tensor_name’: name of the tensor.

  Examples for `torch.nn.Module`s: ‘conv1.weight’,’fc2.bias’.

  • ’round’: 0-based number of round corresponding to this

    tensor.

  • ’tags’: tuple of tensor tags. Tags that can appear:

    • ’model’ indicates that the tensor is a model parameter.

    • ’trained’ indicates that tensor is a part of a training

      result. These tensors are passed to the aggregator node after local learning.

    • ’aggregated’ indicates that tensor is a result of

      aggregation. These tensors are sent to collaborators for the next round.

    • ’delta’ indicates that value is a difference between

      rounds for a specific tensor.

    also one of the tags is a collaborator name if it corresponds to a result of a local task.

  • ’nparray’: value of the tensor.

• tensor_name – name of the tensor

• fl_round – round number

• tags – tuple of tags for this tensor

Returns:

np.ndarray – aggregated tensor

class openfl.interface.aggregation_functions.YogiAdaptiveAggregation(*args, **kwargs)

Yogi adaptive Federated Aggregation funtcion.

adagrad_adaptive_aggregation	Adagrad adaptive aggregation module.
adam_adaptive_aggregation	Adam adaptive aggregation module.
core
experimental
fedcurv_weighted_average	FedCurv Aggregation function module.
geometric_median	Geometric median module.
median	Median module.
weighted_average	Federated averaging module.
yogi_adaptive_aggregation	Yogi adaptive aggregation module.