Multipath Cross Graph Convolution for Knowledge Representation Learning

Table 3

Comparison of the overall effect of multiple baseline models on different datasets.

Dataset		YAGO43kET				WN18RR				FB15kET
Metrics		MRR	H_@1	H_@3	H_@10	MRR	H_@1	H_@3	H_@10	MRR	H_@1	H_@3	H_@10
Methods	TransE	0.21	12.63	23.24	38.93	0.14	8.14	13.27	19.56	0.45	31.51	51.45	73.93
	TransE-ET	0.18	9.19	19.41	35.58	0.16	8.32	14.31	21.22	0.46	33.56	52.96	71.16
	TransR	0.19	10.23	19.97	36.75	0.16	8.44	17.92	26.71	0.47	34.63	53.67	72.02
	ETE	0.23	13.73	26.28	42.18	0.18	9.12	18.21	27.13	0.50	38.51	55.33	72.93
	PTransE	0.24	13.74	26.36	42.33	0.20	10.38	20.32	29.33	0.53	39.87	56.47	73.51
	ConnectE-(E2T+0)	0.25	13.66	26.38	44.60	0.21	11.71	23.31	30.74	0.57	45.82	62.60	80.01
	ConnectE-(E2T + TRT)	0.29	16.13	30.98	47.99	0.23	12.17	23.79	31.03	0.59	49.611	64.69	80.03
	TransC-GCN _AdaGrad	0.29 ± .32	17.12 ± .03	31.33 ± .21	48.72 ± .02	0.23 ± .21	13.76 ± .01	28.35 ± .13	33.07 ± .08	0.61 ± .03	49.47 ± .76	65.25 ± .44	81.02 ± .41
	TransC-GCN _SGD	0.29 ± .26	17.06 ± .17	30.95 ± .36	48.98 ± .00	0.24 ± .32	13.98 ± .36	29.18 ± .27	43.53 ± .03	0.61 ± .37	49.53 ± .40	65.39 ± .68	81.33 ± .32