Table 10.

Top-1 combinations for the MobileNet in the 15 optimization iterations.

#	Parameters optimizer	Batch size	Dropout ratio	TL learn ratio	Loss	Accuracy	F1	Precision	Recall	AUC	WSM
1	AdaDelta	64	0.59	90%	0.6718	79.70%	79.70%	79.70%	79.70%	0.8643	77.78%
2	AdaMax	32	0.6	100%	0.5892	90.57%	90.57%	90.57%	90.57%	0.9482	88.39%
3	AdaMax	32	0.6	100%	0.6249	90.96%	90.96%	90.96%	90.96%	0.9476	88.76%
4	AdaGrad	64	0.31	50%	4.1396	54.35%	54.36%	54.35%	54.35%	0.5796	53.01%
5	SGD	64	0.6	100%	0.1050	96.73%	96.73%	96.73%	96.73%	0.9928	94.79%
6	SGD	64	0.6	100%	0.1323	95.96%	95.96%	95.96%	95.96%	0.9901	93.94%
7	SGD	64	0.6	100%	0.0807	97.69%	97.69%	97.69%	97.69%	0.9950	95.87%
8	SGD	64	0.6	100%	0.0538	98.08%	98.08%	98.08%	98.08%	0.9965	96.55%
9	SGD	64	0.6	100%	0.0433	98.75%	98.75%	98.75%	98.75%	0.9983	97.44%
10	SGD	64	0.6	100%	0.0440	98.46%	98.46%	98.46%	98.46%	0.9964	97.13%
11	SGD	64	0.6	100%	0.1060	97.26%	97.26%	97.26%	97.26%	0.9921	95.30%
12	SGD	64	0.6	100%	0.0610	98.17%	98.17%	98.17%	98.17%	0.9970	96.54%
13	SGD	64	0.6	100%	0.0535	98.27%	98.27%	98.27%	98.27%	0.9982	96.75%
14	SGD	64	0.6	100%	0.0524	98.36%	98.37%	98.36%	98.36%	0.9967	96.86%
15	SGD	64	0.6	100%	0.0413	98.51%	98.51%	98.51%	98.51%	0.9979	97.26%