mkbe.md

Motivation: embedding triplet and multimodal data into vector space
Methods: encoders: multimodal data into vector; decoders: generate multi-modal values
Datasets: MovieLens-100k, YAGO-10
Experiments: link prediction, generating text and images