BucketingAggregationStrategy<T>

Models & Types · AiDotNet.FederatedLearning.Aggregators

Implements the Bucketing meta-strategy for Byzantine-robust federated learning.

Inheritance: FederatedLearningComponentBase<T> → AggregationStrategyBase<Dictionary<string, T[]>, T> → ParameterDictionaryAggregationStrategyBase<T> → BucketingAggregationStrategy<T>

For Beginners

Bucketing is not an aggregation rule itself — it is a “wrapper” that improves any existing robust aggregation rule. Before running the inner aggregator, clients are randomly shuffled into equal-sized buckets. Within each bucket the client updates are averaged, producing one “super-update” per bucket. The inner robust aggregator then operates on these super-updates instead of individual client updates.

How It Works

This provably increases the breakdown point (the fraction of adversaries the defense can tolerate) for any sub-quadratic robust aggregator.

Algorithm:

Reference: Karimireddy, S. P., et al. (2022). “Byzantine-Robust Learning on Heterogeneous Datasets via Bucketing.” ICML 2022.

Constructors

Constructor Summary
BucketingAggregationStrategy(ParameterDictionaryAggregationStrategyBase<T> innerStrategy, int numBuckets, int seed) Initializes a new instance of the BucketingAggregationStrategy class.

Properties

Property Summary
ParameterDictionaryAggregationStrategyBase<T> InnerStrategy { get; } Gets the inner robust aggregation strategy.
int NumBuckets { get; } Gets the number of buckets.

Methods

Method Summary
Dictionary<string, T[]> Aggregate(Dictionary<int, Dictionary<string, T[]>> clientModels, Dictionary<int, double> clientWeights)
string GetStrategyName()