Multisource Mobile Transfer Learning Algorithm Based on Dynamic Model Compression

Table 4

Pruning rate, accuracy (%), and decrease point of the algorithm on the image dataset.

Algorithms	A.W.D- > C			A.W.C- > D			A.D.C- > W			W.D.C- > A
	Pruning(%)	Accuracy(%)	Decrease point	Pruning (%)	Accuracy (%)	Decrease point	Pruning (%)	Accuracy (%)	Decrease point	Pruning (%)	Accuracy (%)	Decrease point
DCTN	0	90.48		0	98.7		0	98.31		0	91.22
	30	90.32	0.16	30	98.65	0.05	30	98.28	0.03	30	91.1	0.12
	50	90.19	0.29	50	98.62	0.08	50	98.21	0.1	50	90.97	0.25
	70	90.07	0.41	70	98.49	0.21	70	98.05	0.26	70	90.79	0.43
	90	89.26	1.22	90	97.69	1.01	90	96.75	1.56	90	89.77	1.45
	95	65.85	24.63	95	68.48	30.22	95	68.06	30.25	95	61.59	29.63
MultiSource TrAdaBoost	0	90.32		0	98.42		0	97.89		0	90.62
	30	90.07	0.25	30	98.3	0.12	30	97.81	0.08	30	90.41	0.21
	50	90.01	0.31	50	98.23	0.19	50	97.77	0.12	50	90.28	0.34
	70	89.87	0.45	70	98.08	0.34	70	97.69	0.2	70	90.1	0.52
	90	88.11	2.21	90	97.2	1.22	90	96.41	1.48	90	88.66	1.96
	95	62.01	28.31	95	71.68	26.74	95	69.58	28.31	95	62.7	27.56
M^3SDA	0	92.55		0	99.29		0	99.06		0	94.08
	30	92.37	0.18	30	99.2	0.09	30	99.03	0.03	30	93.97	0.11
	50	92.35	0.2	50	99.15	0.14	50	98.99	0.07	50	93.89	0.19
	70	92.32	0.23	70	99.11	0.18	70	98.93	0.13	70	93.82	0.26
	90	91.29	1.26	90	98.06	1.23	90	97.51	1.55	90	92.07	2.01
	95	65.22	27.33	95	73.92	25.37	95	68.93	30.13	95	64.84	29.24
MMTLDMC	0	91.88		0	99.46		0	99.52		0	95.11
	X	X	X	X	X	X	X	X	X	X	X	X
	50	91.75	0.13	50	99.4	0.06	50	99.49	0.03	50	95.05	0.06
	70	91.67	0.21	70	99.33	0.13	70	99.47	0.05	70	95.01	0.1
	90	90.77	1.11	90	98.48	0.98	90	98.52	1.25	90	94.13	0.98
	95	67.67	24.21	95	75.14	24.32	95	73.21	26.31	95	67.15	27.96