Blind Federated Learning without initial model

Journal of Big Data

Table 20 Blended Blind FL results (constant 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%	75%	0.9722	0.9722	1.0000	1.0000
2	20%	69%	0.8889	0.9167	0.8750	0.9032
3	20%	69%	0.9444	1.0000	1.0000	1.0000
4	20%	86%	0.9444	1.0000	0.9231	1.0000
5	20%	83%	0.9722	1.0000	0.9630	0.9630
Avg.	–	–	0.9444	0.9778	0.9522	0.9732
1	13%	63%	0.3006	0.3196	0.8929	0.9259
2	30%	80%	0.3731	0.4104	0.8261	0.8947
3	21%	87%	0.3384	0.3720	0.9167	0.8947
4	11%	61%	0.3457	0.3765	0.9565	1.0000
5	24%	81%	0.3510	0.3846	0.8391	0.8676
Avg.	–	–	0.3418	0.3726	0.8862	0.9166
1	4%	75%	0.3165	0.3544	0.7333	0.6923
2	5%	68%	0.2949	0.3077	0.7500	0.6250
3	45%	77%	0.3452	0.3665	0.8837	0.8519
4	13%	51%	0.4378	0.4865	0.4286	0.6154
5	32%	85%	0.2640	0.2775	0.9077	0.9333
Avg.	–	–	0.3317	0.3585	0.7407	0.7436
1	1%	25%	0.5625	0.5625	0.0000	0.0000
2	5%	72%	0.3500	0.3750	0.7143	1.0000
3	45%	78%	0.3236	0.3250	0.9506	0.8545
4	14%	53%	0.4891	0.5163	0.9231	0.7059
5	35%	89%	0.2609	0.2737	0.9130	0.9459
Avg.	–	–	0.3972	0.4105	0.7002	0.7013