Communication Efficient Federated Learning

This reposity contains the code to run simulations from the Communication-Efficient Federated Learning for Wireless Edge Intelligence in IoT paper in IEEE IoT journal.

Requirements

• python = 3.7
• tensorflow = 2.1.0
• numpy = 1.17
• bitarray = 1.2.1

Running

To run a non-IID MNIST CE-FedAvg simulation with 10 workers, C = 0.5, E = 1, sparsity =0.6 using the MNIST-2NN model, run

python main.py

The main.py file is easily edited to run experiments with the CIFAR dataset and different FL parameters.