Table 4.

Hyperpameters of the methods.

Models	Hyperpameters
DNN without UV	Learning rate			Hidden layers
	0.001			64-128-32
DNN with UV	Learning rate			Hidden layers
	0.001			64-128-32
$\beta$SER	Learning rate	$\beta$	Encoder hidden layers	Encoder feature dimension	Decoder hidden layers	Dense layers
	0.001	0.8	64-128-64	5	64-128-64	64-128-64
SERCAM	$\beta$SER module	CAM module
		Learning rate	Hidden layers	Gaussian components Num	Sampling Num
	Same as above	0.001	128-32	2	50