Journal of Infrastructure Preservation and Resilience

Table 6 Optimal combination of the hyperparameters for the models developed with Strategy B

From: Modeling retroreflectivity degradation of pavement markings across the US with advanced machine learning algorithms

DT models					ANN models
Models	D	S	L	F	Models	H	N	B	A	O	Lr	α
DT-A	10	18	2	auto	ANN-A	2	(30, 60)	32	ReLU	adam	0.001	0.1
DT-B	10	10	2	auto	ANN-B	2	(30, 60)	64	ReLU	adam	0.001	0.1
DT-C	10	10	2	auto	ANN-C	2	(30, 60)	64	ReLU	adam	0.001	0.1
DT-D	20	2	2	sqrt	ANN-D	2	(30, 60)	32	ReLU	adam	0.001	0.4
DT-E	10	50	20	auto	ANN-E	1	(30,)	32	ReLU	adam	0.001	0.1
DT-F	20	2	2	auto	ANN-F	1	(30,)	32	ReLU	adam	0.001	0.1
DT-G	20	50	2	auto	ANN-G	1	(60,)	64	ReLU	adam	0.001	0.6
DT-H	10	2	2	auto	ANN-H	2	(30, 60)	64	ReLU	adam	0.001	0.6
DT-I	50	2	2	sqrt	ANN-I	2	(30, 60)	64	ReLU	adam	0.001	0.1
DT-J	10	2	10	sqrt	ANN-J	2	(30, 60)	32	ReLU	adam	0.001	0.1
DT-K	4	2	2	auto	ANN-K	1	(30,)	32	ReLU	adam	0.001	0.1

Back to article page