We provide all code in the form of Jupyter source files.

Please ensure the running environment includes the following Python packages in order to run each of the codes.

---------------------------------------------------------------------------------------------------------
networkx
gossipy-dfl
gossipy
ipywidgets
librosa
tensorflow_datasets
accelerate
opencv-contrib-python
opencv-python
visdom
torchsummary
pytorch-accelerated
ffmpeg
jupyterlab
jupyter

---------------------------------------------------------------------------------------------------------

Additional explanations for asymmetric noise model: 

In the CIFAR-10 dataset, data in the automobile class swaps label with that in the truck class (denoted as automobile(Label index is 1)↔truck(Label index is 9)), and cat(Label index is 3)↔dog(Label index is 5); in MNIST, we have 1↔7 and 0↔6; in Fashion-MNIST, we have Tshirt(Label index is 0)↔shirt(Label index is 6) and pullover(Label index is 2)↔coat(Label index is 4).