An Improved Adam Optimization Algorithm Combining Adaptive Coefficients and Composite Gradients Based on Randomized Block Coordinate Descent

Table 7

Training error and test error of six methods.

	Training error
	SGD_BP	AdaGrad_BP	Adam_BP	C-Adam_BP	RBC-Adam_BP	ACGB-Adam_BP
MSE	1.174293	0.51372	0.095824	0.021837	0.101693	0.019725
RMSE	1.083648	0.716743	0.309555	0.147773	0.318893	0.140446
Test error
MSE	SGD_BP	AdaGrad_BP	Adam_BP	C-Adam_BP	RBC-Adam_BP	ACGB-Adam_BP
	1.267104	0.566655	0.163862	0.028447	0.155603	0.022442
RMSE	1.125657	0.752772	0.404799	0.168662	0.394465	0.149807

The minimum values of MSE and RMSE in Table 7 is shown in bold.