notesum.ai

Published at November 21

REFOL: Resource-Efficient Federated Online Learning for Traffic Flow Forecasting

Released Date: November 21, 2024

Authors: Qingxiang Liu, Sheng Sun, Yuxuan Liang, Xiaolong Xu, Min Liu, Muhammad Bilal, Yuwei Wang, Xujing Li, Yu Zheng

Arxiv: http://arxiv.org/abs/2411.14046v1

Refer to caption

Methods	PEMS-BAY						METR-LA
Methods	5min ( $F=1$ )		30min ( $F=6$ )		1h ( $F=12$ )		5min ( $F=1$ )		30min ( $F=6$ )		1h ( $F=12$ )
	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE
SVR	2.01	1.09	3.74	1.79	4.80	2.28	6.43	3.46	8.41	4.34	9.61	5.00
GRU	2.02	0.99	3.93	1.77	5.50	2.42	5.56	2.91	8.13	4.17	9.41	4.90
DCRNN	1.65	0.93	4.79	2.09	6.19	2.71	4.30	3.57	7.09	4.61	8.71	5.33
STGCN	1.46	0.86	3.62	1.73	4.43	2.26	5.92	3.82	8.60	4.62	10.66	5.92
MegaCRN	1.61	0.91	4.42	1.94	5.33	2.33	4.24	3.54	6.77	4.51	8.11	5.09
FedAvg	1.80	1.02	3.72	1.78	5.25	2.41	5.84	3.32	8.16	4.45	9.52	5.22
FedGRU	1.82	0.97	4.72	2.27	6.52	3.37	5.71	2.99	9.23	5.00	11.03	6.56
FCGCN	9.25	4.82	9.45	4.97	9.36	5.04	15.37	9.69	15.45	10.04	15.16	9.83
FASTGNN	5.32	2.82	6.06	3.20	6.84	3.76	11.73	7.27	12.98	8.23	12.82	8.16
CNFGNN	1.82	1.04	3.60	1.73	5.13	2.64	5.82	3.25	8.03	4.24	9.38	5.12
FedGTP	1.75	1.00	4.59	2.06	6.10	2.74	6.11	2.96	9.34	4.40	11.77	5.66
pFedCTP	5.29	2.84	6.49	3.40	7.43	3.92	8.88	5.21	10.56	6.02	12.08	7.15
REFOL	1.00	1.00	1.86	1.61	2.44	2.08	3.29	3.29	4.87	4.02	5.29	4.21