Blind Federated Learning without initial model

Journal of Big Data

Table 27 Blended Blind FL results (Precision-based weights)—adult dataset

Agent	Size	% 1s	Accuracy	Accuracy	Precision	Precision
Agent	Size	% 1s	Pre-FL	Post-FL	Pre-FL	Post-FL
1	20%	72%	0.7777	0.8888	0.8214	0.8928
2	20%	72%	0.9166	0.9444	0.9600	1.0000
3	20%	81%	0.8888	0.9444	0.9629	1.0000
4	20%	69%	0.8611	0.9166	0.9166	0.9200
5	20%	83%	0.9166	0.9166	0.9354	0.9354
Avg.	–	–	0.8722	0.9222	0.9193	0.9497
1	13%	63%	0.4680	0.5106	0.5710	0.6667
2	30%	80%	0.3296	0.3296	1.0000	1.0000
3	21%	87%	0.2631	0.2894	1.0000	1.0000
4	11%	61%	0.5454	0.6136	0.6667	0.6000
5	24%	81%	0.3619	0.3619	1.0000	0.8000
Avg.	–	–	0.3936	0.4210	0.8476	0.8133
1	4%	75%	0.3888	0.3888	1.0000	1.0000
2	5%	68%	0.2380	0.2857	0.5000	0.0000
3	45%	77%	0.3293	0.3652	1.0000	0.7222
4	13%	51%	0.4772	0.5454	0.0000	0.5000
5	32%	85%	0.3214	0.3482	0.9333	0.8333
Avg.	–	–	0.3510	0.3867	0.6867	0.6111
1	1%	25%	0.6250	0.6250	0.0000	0.0000
2	5%	72%	0.4117	0.4705	0.7142	0.8000
3	45%	78%	0.3488	0.3604	0.8437	0.7222
4	14%	53%	0.5531	0.5531	0.6666	0.5000
5	35%	89%	0.1949	0.2372	1.0000	0.8750
Avg.	–	–	0.4267	0.4493	0.6449	0.5794