notesum.ai

Published at November 6

Fully Hyperbolic Rotation for Knowledge Graph Embedding

cs.AI

cs.LG

Released Date: November 6, 2024

Authors: Qiuyu Liang¹, Weihua Wang¹, Feilong Bao¹, Guanglai Gao¹

Aff.: ¹College of Computer Science, Inner Mongolia University, Hohhot, China

Arxiv: http://arxiv.org/abs/2411.03622v1

Refer to caption

Space	Model	FB15k-237					WN18RR
Space	Model	MRR	H@1	H@3	H@10		MRR	H@1	H@3	H@10
$\mathbb{E}$	MuRE^♣	0.313	0.226	0.340	0.489		0.458	0.421	0.471	0.525
	RotE	0.307	0.220	0.337	0.482		0.463	0.426	0.477	0.529
	Rot2L	0.326	0.237	-	0.503		0.475	0.434	-	0.554
	SAttLE-Tucker	0.340	0.252	0.372	0.513		0.454	0.414	0.474	0.527
$\mathbb{C}$	RotatE^♣	0.290	0.208	0.316	0.458		0.378	0.330	0.417	0.491
$\mathbb{C}$	ComplEx-N3^♣	0.294	0.211	0.322	0.463		0.420	0.390	0.420	0.460
$\mathbb{H}$	MuRP^♣	0.323	0.235	0.353	0.501		0.465	0.420	0.484	0.544
	RotH	0.314	0.223	0.346	0.497		0.472	0.428	0.490	0.553
	FFTRotH	0.319	0.228	0.352	0.500		0.484	0.437	0.502	0.572
	HYBONET	0.334	0.244	0.365	0.516		0.489	0.455	0.503	0.553
	UltraE	0.334	0.243	0.360	0.510		0.488	0.440	0.503	0.558
	FHRE(ours)	0.345	0.255	0.375	0.528		0.494	0.458	0.510	0.563