. 2019 Apr 8;32(6):899–918. doi: 10.1007/s10278-019-00196-1

Table 5.

Hyper-parameters

Datasets	Batch size		Number of epochs	Regularization	Optimization method	Learning rate	Learning rate decay	Beta-1	Beta-2	Epsilon
Datasets	Pre-training	Fine-tuning	Number of epochs	Regularization	Optimization method	Learning rate	Learning rate decay	Beta-1	Beta-2	Epsilon
ABIDE I	30	22	50	0.01	Adam	0.01	0	0.9	0.999	1e-08
ABIDE I		22			Adamax	0.002
ABIDE II		20			Adam	0.01
ABIDE II		20			Adamax	0.002
ABIDE I and ABIDE II		20			Adam	0.01
ABIDE I and ABIDE II		20			Adamax	0.002