Multiple Run Ensemble Learning with Low-Dimensional KGEs (Research Colloquium presentation by Cosimo Gregucci)

a year ago

Presentation of the related work paper Multiple Run Ensemble Learning with Low-Dimensional KGEs

Cosimo Gregucci- ESR 10, University of Stuttgart

During the colloquium, ESR 10 presented the paper “Multiple Run Ensemble Learning with Low-Dimensional KGEs”, which constitutes a relevant state-of-the-art work for his current research.

The main intuition of the paper is that, instead of training large-sized embedding models, which would increase the training cost and the risk of overfitting, it’s possible to obtain even better results by combining multiple runs of a low-dimension embedding model. Besides, the paper shows that with this ensemble technique, the generalization capabilities of the embedding model slightly increase.

The slides of the presentation can be found here.