Sai Praneeth Karimireddy

Feature request: Convert standard dataset into a federated dataset

5

Synthetic federated datasets can constructed from standard centralized ones by artificially splitting them among clients. This is usually done using a Dirichlet distribution (e.g. [Hsu et al. 2019](https://arxiv.org/pdf/1909.06335.pdf)). Such synthetic...

Implementing SCAFFOLD

2

It might be a good idea to have an implementation of SCAFFOLD as well in the algorithms. I think this can be done by modifying the existing Mime implementation.

Centralized (server-only) algorithm

1

Add functionality which a single update step on the clients using their entire batch. This replicates running centralized algorithms on federated datasets.

Some of the notation is different in the paper and in the code repo. - Memory --> error feedback - Scaled SignSGD --> SignSGD - Signum --> SignSGDm - SignSGD...

It is probably a good idea to include a "standard" synthetic baseline (say EMNIST or MNIST). We can use the performance on these well known datasets to convince the readers/reviewers...

Modified csrankings-s.csv to add myself (Sai Praneeth Karimireddy). I am a new tenure track assistant professor in the CS department at USC. # Contributing to CSrankings Thanks for contributing to...