A High-Precision Classification Method of Mammary Cancer Based on Improved DenseNet Driven by an Attention Mechanism

Table 3

Average accuracy obtained by adjusting different parameters.

Image size	An optimization method	Vector	Batch size	The number of iterations	Average accuracy (%)
	Adam	0.001	32	500	99.06
	SGD	0.001	32	500	97.69
	Adam	0.001	32	500	98.75
	Adam	0.001	8	500	97.88
	Adam	0.001	8	500	98.12
	Adam	0.001	16	500	98.92
	Adam	0.0001	32	500	99.48
	Adam	0.0001	32	300	99.05
	Adam	0.0001	32	1000	98.99