fedbase 0.2.2

An easy, silly, DIY Federated Learning framework with many baselines for individual researchers.

FedBase

An easy, silly, DIY Federated Learning framework with many baselines for individual researchers.

Three steps to achieve FedAvg!

Design philosophy

1. Dataset
   1. Dataset partition
      1. IID
      2. Non-IID
         1. Dirichlet distribution
         2. N classes
         3. ...
      3. Fake data
      4. ...
   2. Batch_size
2. Node
   1. Local dataset
   2. Model
   3. Objective
   4. Optimizer
   5. Local update
   6. Test
3. Server
   1. Model
   2. Aggregate
   3. Distribute
4. Server & Node
   1. Topology
   2. Client sampling
   3. Exchange file
5. Baselines
   1. Global
   2. Local
   3. FedAvg
6. Visualization

How to develop your own FL with fedbase?

Baselines

1. Global training
2. FedAvg

To be continued...