Michael Galkin

Adding Relation Prediction Loss to LP training

3

UCL & FAIR recently showed that adding Relation Prediction loss to standard LCWA / SLCWA can slightly increase the link prediction performance. Implementation seems quite straightforward Paper: https://openreview.net/pdf?id=Qa3uS3H7-Le Code: https://github.com/facebookresearch/ssl-relation-prediction

🦓 🐒 NeurIPS'21: RotPro

2

NeurIPS'21 brought a bunch of KGE models. This is for RotPro. ### Publication Link [Rot-Pro: Modeling Transitivity by Projection in Knowledge Graph Embedding](https://arxiv.org/pdf/2110.14450.pdf) ### Reference Implementation Code: https://github.com/tewiSong/Rot-Pro ### Relevance...

🦓 🐒 NeurIPS'21: ConE + 2 new datasets (GO21 + DDB14)

3

NeurIPS'21 brought a bunch of KGE models. This issue is for ConE. ### Publication Link [Modeling Heterogeneous Hierarchies with Relation-specific Hyperbolic Cones](https://papers.nips.cc/paper/2021/file/662a2e96162905620397b19c9d249781-Paper.pdf) ### Reference Implementation Code: https://github.com/snap-stanford/ConE ### Relevance Published...

🐍 🌟 Hyper-Relational Statement Factory

1

This PR introduces a series of efforts on integrating hyper-relational graphs (aka RDF* or LPG) into PyKEEN with factories, datasets, and models. Here is the adaptation of the [code](https://github.com/mali-git/hyper_relational_ilp) of...

`litellm_get_search_query` does not use a local llm config

2

In the standard `ask` pipeline, the function `litellm_get_search_query` does not employ custom llm settings config available in `query.settings.llm_config`: https://github.com/Future-House/paper-qa/blob/e49f1af67e5463ee5188cbe20bdb9747f14db1f9/paperqa/agents/main.py#L178-L180 This leads to creating a default LiteLLMModel instance which by default...