A split-federated learning and edge-cloud based efficient and privacy-preserving large-scale item recommendation model

Journal of Cloud Computing

Advances, Systems and Applications

Table 4 The efficiency comparison results of FedDSSM, SpFedRec, and SpFedRec-SENet on MovieLens and Adressa

	MovieLens			Adressa
	FedDSSM	SpFedRec	SpFedRec-SENet	FedDSSM	SpFedRec	SpFedRec-SENet
Device Training	1170 ms	228 ms	449 ms	2570 ms	324 ms	681 ms
Inferring Time	4090 ms	55 ms	55 ms	19120 ms	250 ms	250 ms
Total Time	5260 ms	283 ms	471 ms	21690 ms	574 ms	931 ms
< Improvement >	-	18 ×	11 ×	-	38 ×	23 ×
Training Comm	5.42 MB	1.05 MB	1.73 MB	13.68 MB	2.18 MB	4.40 MB
Inferring Comm	0.14 MB	0.05 MB	0.5 MB	557.80 MB	7.19 MB	7.19 MB
Total Comm	5.56 MB	1.10 MB	1.78 MB	571.48 MB	9.37 MB	11.59 MB
< Improvement >	-	5 ×	3 ×	-	60 ×	49 ×