Table 4.

Best hyper-parameters of the MSI-MFNet and DNet classification models.

DataSet	Parameters	Without Data Augmentation		Data Augmentation
		MSI-MFNet	DNet	MSI-MFNet	DNet
ICIAR2018	Number of layers	205	201	205	201
	Epochs	32	32	64	64
	Learning Rate	${10}^{-3}$	${10}^{-2}$	${10}^{-4}$	${10}^{-3}$
	Batch Size	Adam	Adam	Adam	Adam
	Dropout Rate	0.5	0.6	0.5	0.6
BreakHis	Number of layers	205	201	205	201
	Epochs	32	64	64	32
	Learning Rate	${10}^{-4}$	${10}^{-3}$	${10}^{-4}$	${10}^{-3}$
	Batch Size	Adam	Adam	Adam	Adam
	Dropout Rate	0.5	0.5	0.5	0.6