EURASIP Journal on Embedded Systems

Table 3 RMSE comparison of different architectures and training algorithms

From: Modeling of smartphones’ power using neural networks

Training function
	FFBP		CFFBP		FFBPTD
	n = 10	n = 20	n = 10	n = 20	n = 10	n = 20
trainbfg	0.209797	0.206758	0.216204	0.199763	0.632268	0.632285
trainbr	0.193328	0.175663	0.184413	0.174294	0.632267	0.632267
traincgb	0.218694	0.217632	0.21954	0.217823	0.632268	0.632271
traincgf	0.229647	0.226134	0.227723	0.213663	0.632267	0.632267
traincgp	0.225715	0.219247	0.227335	0.217196	0.632269	0.632268
traingd	0.353894	1.192112	0.336454	5.590423	0.632267	0.63227
traingdm	0.966383	0.845086	3.510545	5.550865	0.883342	0.654927
traingda	0.332997	0.39005	0.454985	0.482595	0.632267	0.632568
traingdx	0.286482	0.284275	0.306612	0.317421	0.632344	0.632833
trainlm	0.188031	0.176105	0.190761	0.180157	0.632267	0.632283
trainoss	0.233728	0.236423	0.223991	0.234473	0.632361	0.632269
trainrp	0.231186	0.239729	0.242783	0.247439	0.632274	0.806998
trainscg	0.22851	0.229321	0.232129	0.218688	0.632267	0.632268

Back to article page