Representation Learning of Knowledge Graphs with Embedding Subspaces

Table 4

Zero-shot scenario results of sub-ProjE with different training set sizes.

		Mean rank		HITs@10 (%)
Training set size	Dataset	Raw	Filtered	Raw	Filtered
Train = 5000	FB15K-237	767	769	12.2	12.2
	FB15K	748	746	12.4	12.5
Train = 10000	FB15K-237	699	698	13.0	13.1
	FB15K	669	667	14.4	14.4
Train = all	FB15K-237	524	523	16.3	16.4
	FB15K	531	529	17.4	17.5