FederatedUnlearningOptions
Options & Configuration · AiDotNet.Models.Options
Configuration options for federated unlearning (right to be forgotten).
For Beginners
These options control how the system removes a client’s contribution from the trained model when they exercise their right to be forgotten (GDPR, CCPA, LGPD).
Constructors
| Constructor | Summary |
|---|---|
FederatedUnlearningOptions() |
Properties
| Property | Summary |
|---|---|
UnlearningMethod Method { get; set; } |
Gets or sets the unlearning method. |
bool VerificationEnabled { get; set; } |
Gets or sets whether to verify unlearning correctness. |
double InfluenceTolerance { get; set; } |
Gets or sets the tolerance for influence function convergence. |
double NoiseScale { get; set; } |
Gets or sets the noise scale for diffusive noise unlearning. |
double UnlearningLearningRate { get; set; } |
Gets or sets the learning rate for gradient ascent unlearning. |
double VerificationThreshold { get; set; } |
Gets or sets the verification threshold: maximum cosine similarity between the unlearned model and a model trained with the target client that is considered acceptable. |
int MaxInfluenceIterations { get; set; } |
Gets or sets the maximum number of iterations for influence function computation. |
int MaxUnlearningEpochs { get; set; } |
Gets or sets the maximum epochs for approximate unlearning methods. |