This folder contains the code to reproduce the results presented in “Communication-Efficient Distributionally Robust Decentralized Learning” (https://openreview.net/forum?id=tnRRHzZPMq).
Every sub-folder corresponds to one of the three experimental setups and it contains the implementations of AD-GDA, DRFA and DR-DSGD.
To run in sequence all experimental runs of AD-GDA, run script.py.
To run DRFA and DR-DSGD, run network.py.
To run the experiments on COOS7 is necessary to first download the dataset from https://zenodo.org/record/3355599#.YzrxGy8Rr5g.
Contact: zecchin@eurecom.fr