Testing with different select ratios

[nclabteam]

Firstly thanks for sharing the code.
I am trying to repoduce the results from your paper, if i run the experiments with the default values mentioned in your paper and in args.py, i am able to reproduce the results (20 nodes and select ratio 1.0). But when i changed the select ratio(Client partitcipation ratio) to 0.1 and increase the node num to 100, the accuracy of fedlaw is stuck at 10%.
During your experiments have you encoutered similar issues, or have you tested the algorithm with different select ratios.
And also please clarify the actual select ratio used to produce the results that are mentioned in the paper.

thanks in advance