RSA: Byzantine-Robust Stochastic Aggregation Methods for Distributed Learning from Heterogeneous Datasets

Simulating code of paper https://arxiv.org/abs/1811.03761.

Files:

data/mnist: We load MNIST data from official website and then save as npy file for python. Especially, the directory of "2" is the heterogeneous case.

RSA-CODE: code of our proposed methods --- RSA. RSA-iid.py is about the i.i.d. case, RSA-hete.py is about the heterogeneous case.

geomed: the simulating code of Geometric median algorithm.

krum: the simulating code of Krum algorithm.

median: the simulating code of median algorithm.