notesum.ai

Published at October 23

Multi-Continental Healthcare Modelling Using Blockchain-Enabled Federated Learning

cs.LG

cs.AI

Released Date: October 23, 2024

Authors: Rui Sun¹, Zhipeng Wang², Hengrui Zhang³, Ming Jiang⁴, Yizhe Wen⁵, Jiqun Zhang⁶, Jiahao Sun⁵, Shuoying Zhang⁵, Erwu Liu⁴, Kezhi Li⁷

Aff.: ¹School of Computing, Newcastle University, Newcastle upon Tyne, United Kingdom; ²Department of Computing, Imperial College London, London, United Kingdom; ³Institute of Health Informatics, University College London, London, United Kingdom; ⁴Shanghai Research Institute for Intelligent Autonomous Systems, Tongji University, Shanghai, China; ⁵FLock.io, London, United Kingdom; ⁶College of Electronics and Information Engineering, Tongji University, Shanghai, China; ⁷University College London Hospital, London, United Kingdom

Arxiv: https://arxiv.org/abs/2410.17933v1

RMSE Loss	Selected Current Patients (in ID)							Unseen Patients (in ID)
Method	4	7	13	19	23	Avg	$\Delta_{\text{avg}}$	26	27	28	29	30	Avg	$\Delta_{\text{avg}}$
H1Single	9.243	12.493	11.731	9.058	12.018	10.909	-1.555	11.149	14.810	9.233	9.375	10.012	10.916	-1.642
H2Single	9.874	12.026	12.052	9.231	12.243	11.085	-1.704	10.872	13.609	9.544	9.466	9.970	10.692	-1.418
H3Single	9.342	13.571	11.550	9.324	11.555	11.068	-1.714	10.882	15.769	9.189	9.865	10.273	11.196	-1.922
H4Single	10.180	12.327	11.400	9.341	11.461	10.942	-1.588	10.049	13.990	9.710	9.630	10.388	10.753	-1.479
H5Single	9.674	13.013	11.501	8.946	11.540	10.935	-1.581	10.419	15.212	9.592	9.450	10.469	11.028	-1.754
FedAvg w/ mal	100.608	128.644	113.247	102.614	111.206	111.263	-101.909	100.372	133.070	109.266	102.125	91.790	107.35	-98.076
TotalCentral w/ mal	10.080	12.555	11.836	9.210	12.087	11.154	-1.800	10.682	17.531	9.973	9.639	10.576	11.680	-2.406
TotalCentral	8.740	10.541	10.074	8.202	10.247	9.561	-0.207	9.037	12.119	8.370	8.464	9.100	9.418	-0.144
MCGP w/ mal	8.642	10.936	10.049	8.213	10.273	9.623	-0.269	8.939	13.593	8.386	8.498	8.912	9.666	-0.392
MCGP(ours)	8.650	10.475	9.707	8.094	9.844	9.354	–	8.660	12.621	8.136	8.224	8.728	9.274	–
MARD Loss	Selected Current Patients							Unseen Patients
Method	4	7	13	19	23	Avg	$\Delta_{\text{avg}}$	26	27	28	29	30	Avg	$\Delta_{\text{avg}}$
H1Single	6.538	5.822	5.848	5.340	6.366	5.983	-0.751	6.737	6.107	5.265	5.719	8.144	6.394	-0.959
H2Single	6.799	5.672	5.973	5.420	6.387	6.050	-0.818	6.671	5.782	5.390	5.733	7.879	6.291	-0.856
H3Single	6.574	5.865	5.898	5.477	6.342	6.031	-0.799	6.671	6.138	5.234	5.979	8.232	6.451	-1.016
H4Single	7.005	5.765	5.834	5.389	6.239	6.046	-0.814	6.373	5.822	5.484	5.807	8.209	6.339	-0.904
H5Single	6.796	5.876	5.799	5.298	6.306	6.015	-0.783	6.641	6.188	5.338	5.782	8.481	6.486	-1.051
FedAvg w/ mal	78.861	78.050	78.008	76.768	79.100	78.157	-72.925	79.364	77.981	78.502	76.903	81.648	78.880	-73.445
TotalCentral w/ mal	7.136	5.797	6.087	5.501	6.607	6.226	-0.994	6.742	6.204	5.637	6.009	8.668	6.652	-1.217
TotalCentral	6.144	5.017	5.262	4.915	5.666	5.401	-0.169	5.796	5.128	4.775	5.256	7.239	5.639	-0.204
MCGP w/ mal	5.927	5.111	5.215	4.907	5.611	5.354	-0.122	5.714	5.329	4.753	5.221	7.083	5.620	-0.185
MCGP(ours)	5.955	4.890	5.081	4.798	5.435	5.232	–	5.539	5.069	4.602	5.061	6.904	5.435	–