Privacy Shield: A System for Edge Computing using Asynchronous Federated Learning

Table 3

Performance of the proposed system and LAG for different values of Ω.

Compression algorithm Ω	Privacy shield		LAG
	0.2	0.22
Accr (training set)	93.68%	94.23%	89.9%
Accr (test set)	91.98%	92.35%	89.13%
C-ratio	6.33%	8.77%	5.11%
CBI (c1 = 0.8, c2 = 0.12)	0.9333	0.9206	0.9274
CBI (c1 = 0.10, c2 = 0.10)	0.9325	0.9226	0.922
CBI (c1 = 0.12, c2 = 0.8)	0.9316	0.9247	0.9167