VerticalPartyClient<T>
Models & Types · AiDotNet.FederatedLearning.Vertical
Represents a feature-holding party in vertical federated learning.
Inheritance: FederatedLearningComponentBase<T> → VerticalPartyClient<T>
Implements: IVerticalParty<T>
For Beginners
This is one of the parties in a VFL setup that holds some features (columns) of the dataset but NOT the labels (prediction targets). For example, a bank that has income and credit score data but doesn’t know patient outcomes.
How It Works
The party runs a local “bottom model” (a small neural network) on its features to produce an embedding (compressed representation) that is sent to the coordinator. During backpropagation, the party receives gradients for its embedding and uses them to update its local model.
Privacy guarantee: The raw features never leave the party. Only the embedding (and its gradients) cross party boundaries. The embedding is a lossy compression that makes it difficult to reconstruct the original features.
Constructors
| Constructor | Summary |
|---|---|
VerticalPartyClient(string partyId, Tensor<T> localData, IReadOnlyList<string> entityIds, int embeddingDimension, int hiddenDimension, int? seed) |
Initializes a new instance of VerticalPartyClient. |
Properties
| Property | Summary |
|---|---|
bool IsLabelHolder { get; } |
|
int EmbeddingDimension { get; } |
|
int FeatureCount { get; } |
|
string PartyId { get; } |
Methods
| Method | Summary |
|---|---|
IReadOnlyList<Tensor<T>> GetParameters() |
|
IReadOnlyList<string> GetEntityIds() |
|
Tensor<T> ComputeForward(IReadOnlyList<int> alignedIndices) |
|
void ApplyBackward(Tensor<T> gradients, double learningRate) |
|
void SetParameters(IReadOnlyList<Tensor<T>> parameters) |