nvflare.app_common.statistics.numeric_stats module

accumulate_hists(metrics: Dict[str, Dict[str, Histogram]], global_hists: Dict[str, Dict[str, Histogram]], precision: int = 4) → Dict[str, Dict[str, Histogram]]

accumulate_metrics(metrics: dict, global_metrics: dict, precision: int) → dict

bins_to_dict(bins: List[Bin]) → Dict[BinRange, float]

filter_numeric_features(ds_features: Dict[str, List[Feature]]) → Dict[str, List[Feature]]

get_global_feature_data_types(client_feature_dts: Dict[str, Dict[str, List[Feature]]]) → Dict[str, Dict[str, DataType]]

get_global_stats(global_metrics: dict, client_metrics: dict, metric_task: str, statistic_configs: Dict[str, dict], precision: int = 4) → dict

get_means(sums: dict, counts: dict, precision: int = 4) → dict

get_min_or_max_values(metrics: dict, global_metrics: dict, fn2, precision: int = 4) → dict

Use 2 argument function to calculate fn2(global, client), for example, min or max.

Note

The global min/max values are min/max of all clients and all datasets.

Parameters:

• metrics – client’s metric

• global_metrics – global metrics

• fn2 – two-argument function such as min or max

• precision – decimal number precision

Returns: Dict[dataset, Dict[feature, int]]