Blind Federated Learning without initial model

Journal of Big Data

Table 26 Blind FL results (Precision-based weights, slope 5, sigmoid)—adult dataset

Agent	Size	% 1s	Accuracy	Accuracy	Precision	Precision
Agent	Size	% 1s	Pre-FL	Post-FL	Pre-FL	Post-FL
1	20%	67%	0.9167	0.9722	0.9630	0.9655
2	20%	61%	0.9722	1.0000	1.0000	0.9231
3	20%	72%	0.7500	0.9722	0.7500	0.9643
4	20%	78%	0.8889	0.8889	0.9000	0.9000
5	20%	72%	0.9167	0.9444	0.9231	0.8621
Avg.	–	–	0.8889	0.9556	0.9072	0.9230
1	13%	63%	0.3201	0.3537	0.8718	0.8333
2	30%	80%	0.3621	0.3793	0.8500	0.7143
3	21%	87%	0.3692	0.3754	0.9623	0.9730
4	11%	61%	0.4740	0.4805	0.9722	0.9259
5	24%	81%	0.3255	0.3458	0.8632	0.9111
Avg.	–	–	0.3702	0.3869	0.9039	0.8715
1	4%	75%	0.3117	0.3247	0.8182	0.7000
2	5%	68%	0.2821	0.2821	1.0000	0.6667
3	45%	77%	0.3495	0.3774	0.8532	0.8514
4	13%	51%	0.5161	0.5215	0.8421	0.6429
5	32%	85%	0.2851	0.3050	0.8750	0.8980
Avg.	–	–	0.3489	0.3621	0.8777	0.7518
1	1%	25%	0.5588	0.5882	0.3333	0.0000
2	5%	72%	0.2800	0.2800	1.0000	0.8000
3	45%	78%	0.3645	0.3862	0.8915	0.9506
4	14%	53%	0.4754	0.4809	0.7500	0.4667
5	35%	89%	0.2393	0.2935	0.8025	0.9028
Avg.	–	–	0.3836	0.4057	0.7555	0.6240