Instance-Wise Denoising Autoencoder for High Dimensional Data

Table 7

Transfer learning results on Amazon dataset: comparison results in terms of training time (s).

Task	SVM-SC	CL-KCCA	HeMap	Msda-CCA	HHTL (1 layer)	HHTL (3 layers)	Ours (1 layer)	Ours (3 layers)
EN-B-FR-D	73.25	50.00	49.45	72.96	10.62	14.23	8.31	10.54
EN-B-FR-M	62.40	47.20	50.70	64.29	10.22	15.65	8.23	10.11
EN-B-GE-D	72.55	77.50	66.12	78.23	10.74	14.34	8.71	11.51
EN-B-GE-M	58.95	50.00	48.35	62.52	11.64	13.67	8.53	12.24
EN-B-JP-D	69.50	69.46	49.55	71.95	10.74	14.32	8.91	11.84
EN-B-JP-M	53.15	55.23	52.05	57.17	10.75	15.74	8.29	13.14
EN-D-FR-B	71.35	47.10	48.35	72.32	11.23	14.44	8.57	12.57
EN-D-FR-M	71.65	47.10	49.90	68.87	11.36	16.46	8.90	11.51
EN-D-GE-B	76.40	54.87	50.80	75.32	10.85	14.86	8.87	11.94
EN-D-GE-M	69.05	54.87	50.55	75.69	10.23	14.94	8.66	12.64
EN-D-JP-B	70.00	61.16	50.12	72.60	10.68	13.43	8.60	12.44
EN-D-JP-M	59.20	66.89	49.80	57.75	11.47	16.27	8.94	11.67
EN-M-FR-B	76.20	73.54	50.55	74.23	10.82	14.38	8.25	11.98
EN-M-FR-D	73.65	47.10	50.87	72.76	11.43	14.46	8.16	12.56
EN-M-GE-B	74.50	81.50	49.45	76.82	10.97	13.64	8.75	12.09
EN-M-GE-D	74.60	80.20	50.20	72.28	10.85	14.43	8.69	11.17
EN-B-JP-B	67.85	63.75	48.85	68.83	10.68	16.85	8.91	11.38
EN-B-JP-D	69.35	60.46	48.80	71.06	10.97	14.93	8.50	11.23