Joint Transfer Extreme Learning Machine with Cross-Domain Mean Approximation and Output Weight Alignment

Table 2

Accuracy of different algorithms on USPS + MNIST and Office + Caltech datasets.

Methods\dataset	Nontransfer learning algorithm				Transfer learning algorithm
	1NN	SVM	ELM	SSELM	TCA1	TCA2	JDA1	JDA2	DAELM_S	DAELM_T	TELM-OWA	AELM	CdELM-C	JTELM
USPS vs. MNIST	44.70	35.10	30.70	29.05	50.35	60.20	46.80	53.25	44.36	31.08	59.55	41.65	45.89	62.00
MNIST vs. USPS	65.94	54.69	51.39	43.28	62.28	71.44	57.83	75.94	47.11	60.32	59.37	62.83	42.57	76.39
Average	55.32	44.90	41.05	36.16	56.31	65.82	52.32	64.60	45.73	45.70	59.46	52.24	44.23	69.20
C⟶A (1)	23.70	52.40	50.63	52.92	46.14	44.78	57.93	56.78	50.26	27.85	31.73	52.40	56.28	57.30
C⟶W (2)	25.76	40.68	42.71	45.42	37.97	41.69	48.47	50.85	45.42	31.86	32.20	43.73	50.98	55.59
C⟶D (3)	25.48	42.04	42.04	38.85	45.86	45.22	49.68	50.32	37.58	22.93	33.12	47.77	43.82	52.86
A⟶C (4)	26.00	44.08	43.01	41.76	40.61	39.36	45.68	43.10	42.88	25.25	38.85	40.16	43.17	41.85
A⟶W (5)	29.83	41.69	37.63	35.93	40.00	37.97	41.02	41.36	38.31	19.32	38.64	34.92	40.75	49.49
A⟶D (6)	25.48	40.76	42.68	38.85	31.85	39.49	42.04	41.40	36.94	36.31	43.31	35.67	41.78	49.68
W⟶C (7)	19.86	29.92	32.24	34.28	31.26	31.17	31.43	33.30	34.20	41.90	33.48	30.54	34.44	36.60
W⟶A (8)	22.96	34.13	36.85	37.06	31.94	32.78	37.37	46.03	36.44	46.49	38.01	37.16	39.58	42.28
W⟶D (9)	59.24	87.90	80.25	78.98	89.17	89.17	85.99	85.99	80.89	53.50	77.07	86.62	83.06	82.17
D⟶C (10)	26.27	31.43	30.10	30.37	33.39	31.52	34.91	35.26	32.32	48.08	32.50	28.67	34.44	37.49
D⟶A (11)	28.50	32.36	31.52	32.05	31.21	33.09	38.94	40.40	34.35	51.20	36.23	29.54	39.54	43.63
D⟶W (12)	63.39	73.56	71.86	80.00	87.46	89.49	81.69	83.05	72.88	54.24	78.64	83.73	84.34	83.73
Average	31.37	45.91	45.13	45.54	45.57	46.31	49.60	50.65	45.21	38.24	42.82	45.91	49.35	52.72
Total average	34.79	45.77	44.54	44.20	47.11	49.10	49.98	52.65	45.28	39.31	45.19	46.81	48.62	55.08

Bold values indicate that the value is the best result of the row in which it is located.