Table 2.

Comparison of test results with different parameter configurations.

Learning rate	Epoch	Batch size	Optimizer	Test loss	Test accuracy
0.001	50	64	Adam	0.9687	96.33%
0.001	100	32	Adam	0.0685	97.36%
0.001	150	16	Adam	0.0744	97.79%
0.0001	50	16	Adam	0.0865	97.78%
0.0001	150	32	Adam	0.6556	97.82%
0.0001	100	64	Adam	0.0265	98.05%
0.0002	50	64	Adam	0.1243	98.01%
0.0002	100	32	Adam	0.0434	97.73%
0.0002	150	16	Adam	0.0792	97.56%
0.001	50	64	Adagrad	0.9699	95.34%
0.001	100	32	Adagrad	0.0768	96.27%
0.001	150	16	Adagrad	0.0868	96.52%
0.0001	50	16	Adagrad	0.0978	96.98%
0.0001	150	32	Adagrad	0.7686	96.92%
0.0001	100	64	Adagrad	0.0357	97.85%
0.0002	50	64	Adagrad	0.2126	97.02%
0.0002	100	32	Adagrad	0.0567	95.56%
0.0002	150	16	Adagrad	0.0876	96.88%

The bold values are used to highlight that they are optimal compared to other numerical combinations. They are also the final parameter settings used in the model we proposed.