[Retracted] Gradient Descent Optimization in Deep Learning Model Training Based on Multistage and Method Combination Strategy

Table 15

Performance of the proposed method, Adam with decay.

	ResNet-20 on Cafri-10		LSTM on IMDB
	Val-loss	Val-acc	Val-loss	Val-acc
(Adam + d) + SGD	0.5980	0.8513	0.9323	0.8143
(Adam + d) + (SGD + d)	0.5959	0.8528	0.9291	0.8150
(Adam + d) + (SGD + M)	0.6745	0.8217	1.0951	0.8172
(Adam + d) + (SGD + M + d)	0.6960	0.8232	1.1128	0.8137
(Adam + d) + RMSprop	0.7657	0.8078	1.3052	0.7917
(Adam + d) + (RMSprop + d)	0.7760	0.8100	1.0837	0.8137
(Adam + d) + Adam	1.0136	0.7399	1.1720	0.8048
(Adam + d) + (Adam + d)	0.9641	0.7509	1.2429	0.8060

The bold values represent the best results.