CifarFederatedBenchmarkOptions
Options & Configuration · AiDotNet.Configuration
Configuration options for running CIFAR-based federated benchmark suites.
For Beginners
This suite tests image classification using the well-known CIFAR datasets. You point AiDotNet to the extracted CIFAR folder, and it handles the rest.
How It Works
CIFAR datasets are distributed as binary batch files. AiDotNet loads the dataset from the provided directory, applies an industry-standard synthetic federated partitioning strategy (for example, Dirichlet label skew), and evaluates the model with a structured report.
Constructors
| Constructor | Summary |
|---|---|
CifarFederatedBenchmarkOptions() |
Properties
| Property | Summary |
|---|---|
FederatedPartitioningStrategy PartitioningStrategy { get; set; } |
Gets or sets the dataset-to-client partitioning strategy. |
bool NormalizePixels { get; set; } |
Gets or sets whether pixel values should be normalized to the range [0,1]. |
double DirichletAlpha { get; set; } |
Gets or sets the Dirichlet concentration parameter used when PartitioningStrategy is DirichletLabel. |
int? ClientCount { get; set; } |
Gets or sets the number of federated clients to simulate (null uses defaults). |
int? MaxTestSamples { get; set; } |
Gets or sets an optional maximum number of test samples to load (null loads all available). |
int? MaxTrainSamples { get; set; } |
Gets or sets an optional maximum number of training samples to load (null loads all available). |
int? ShardsPerClient { get; set; } |
Gets or sets how many label shards should be assigned to each client when using ShardByLabel. |
string DataDirectoryPath { get; set; } |
Gets or sets the dataset directory path containing CIFAR binary files. |