NonIIDClientSplitter<T>
Models & Types · AiDotNet.Preprocessing.DataPreparation.Splitting.FederatedLearning
Non-IID (non-Independent and Identically Distributed) client splitter for federated learning.
Inheritance: DataSplitterBase<T> → NonIIDClientSplitter<T>
For Beginners
In real-world federated learning, data on different clients is often NOT identically distributed. For example, users’ photos on different phones reflect their individual preferences. This splitter simulates such heterogeneous distributions.
How It Works
Heterogeneity Types:
- Label skew: Each client has only some classes
- Quantity skew: Clients have different amounts of data
- Feature skew: Clients have different feature distributions
When to Use:
- Realistic federated learning experiments
- Testing robustness to heterogeneous data
- Simulating domain adaptation scenarios
Constructors
| Constructor | Summary |
|---|---|
NonIIDClientSplitter(int numClients, int classesPerClient, double testRatio, bool allowOverlap, bool shuffle, int randomSeed) |
Creates a new Non-IID client splitter. |
Properties
| Property | Summary |
|---|---|
bool RequiresLabels { get; } |
|
int NumSplits { get; } |
|
string Description { get; } |
Methods
| Method | Summary |
|---|---|
DataSplitResult<T> Split(Matrix<T> X, Vector<T> y) |
|
IEnumerable<DataSplitResult<T>> GetSplits(Matrix<T> X, Vector<T> y) |