See the following paper for reference on the metrics which need to be implemented (section 5., especially 5.1.1 and 5.1.2; 5.1.4 may also be of interest, we could treat this as a sort of approx problem to the Yoneda/co-Yoneda embedding; i.e 2 objects are approx equal if they have approx the same maps onto them (Yoneda)/out of them(co-Yoneda)):
https://persagen.com/files/misc/Wang2017Knowledge.pdf
Make necessary adjutments to test our method for knowledge graph embedding on WN18 and BF15K datasets.
See the following repository for reference on the datasets: https://github.com/villmow/datasets_knowledge_embedding
See the following paper for reference on the metrics which need to be implemented (section 5., especially 5.1.1 and 5.1.2; 5.1.4 may also be of interest, we could treat this as a sort of approx problem to the Yoneda/co-Yoneda embedding; i.e 2 objects are approx equal if they have approx the same maps onto them (Yoneda)/out of them(co-Yoneda)): https://persagen.com/files/misc/Wang2017Knowledge.pdf