Journal of Big Data

Table 3 Time required (minutes) to train the two models with the best efficacy

From: A new deep learning architecture with inductive bias balance for transformer oil temperature forecasting

Dataset	Window size	SCINet	SRCNet
ETTh1	24	1.10	0.50
	48	2.14	1.26
	72	2.81	1.90
	96	3.70	1.01
	120	1.75	1.04
ETTh2	24	3.97	3.23
	48	5.23	1.20
	72	5.60	1.57
	96	5.05	1.17
	120	7.24	1.25

Back to article page