Table 5 Compared models and hyperparameter settings for multi-label SE identification.

Model	Hyperparameters	Value
CC	C	1
	Solver	liblinear
	Class weight	balanced
	Max_iter	2000
	Chain order	random (order = None)
ExtraTrees	n_estimators	500
	max_depth	None
	min_samples_split	2
	min_samples_leaf	1
	max_features	sqrt
MMoE	Learning rate	0.001
	Batch size	32
	Max_epochs	100
	Optimizer	Adam
	Dropout	0.1
	Num experts	4
	Expert out dim	32
	Expert hidden dim	64
	Tower hidden dim	64
	Activation functions	ReLU (experts/towers), Sigmoid (output)
MLP	Learning rate	0.001
	Batch size	64
	Hidden dim	64
	Max_epochs	150
	Optimizer	Adam
	Dropout	0.2
	Activation functions	GELU (backbone), Sigmoid (output)
RIAM_Multi	Learning rate	0.001
	Batch size	32
	dl	64
	Max_epochs	200
	Optimizer	Adam
	Dropout	0.1
	Input dimensions	6
	Output dimensions	3
	Activation functions	Sigmoid
	U	2
	N_x	2

Quick links

Search