SelecEval Documentation

Uniform

SelecEval Documentation

Introduction
Configuration
Installation
Thesis
Algorithms
Algorithms
- ActiveFL
- CEP
- FedCS
- Pow-D
- Random
Data
Data
- Adding new data distributions
- Datasets
  Datasets
- Feature distributions
  Feature distributions
  - Gaussian
- Label distributions
  Label distributions
- Quantity distributions
  Quantity distributions
  - Dirichlet
  - Uniform
Examples
Examples
- Example Descriptions
- Example 1
- Example 2
- Paper
  Paper
Module docs
Module docs
- __main__
- Client
- DataHandler
- Models
- Selection
- Strategy
- Utility
- Validation
Simulation
Simulation
Validation
Validation
- Configuration
- Output
- Advanced analytics
  Advanced analytics

Uniform

When using the uniform distribution, every client is allocated the same number of samples of each class. This leads to better training convergence, as the data is now i.i.d., however, this is not realistic for a real-world FL scenario.