Table 1 Comparision of RQSGD on MNIST with different \(\boldsymbol{\alpha }\).
From: Two-layer accumulated quantized compression for communication-efficient federated learning: TLAQC
∝ | Accuracy |
|---|---|
1 | 98.32 |
0.9 | 98.35 |
0.8 | 98.36 |
0.7 | 98.29 |
0.6 | 98.28 |
0.5 | 93.31 |
0.4 | 98.30 |
0.3 | 98.30 |
0.2 | 98.28 |
0.1 | 98.27 |