Integrating new models

[adelmemariani]

We plan to integrate the following models into the dice-embeddings. Please feel free to add other models to this list :)

GNN-based:
RGCN [1], CompGCN[2]

Rule-based:
RNNLogic[3], NCRL[4]

[1] Modeling Relational Data with Graph Convolutional Networks (https://link.springer.com/chapter/10.1007/978-3-319-93417-4_38)
[2] Composition-based Multi-Relational Graph Convolutional Networks (https://arxiv.org/pdf/1911.03082)
[3] RNNLogic: Learning Logic Rules for Reasoning on Knowledge Graphs (https://arxiv.org/abs/2010.04029)
[4] Neural Compositional Rule Learning for Knowledge Graph Reasoning (https://arxiv.org/abs/2303.03581)

[adelmemariani]

GNNs Meet Sequence Models Along the Shortest-Path: an Expressive Method for Link Prediction (https://arxiv.org/abs/2507.07138)

PROXI: Challenging the GNNs for Link Prediction (PROXI: Challenging the GNNs for Link Prediction)

Boosting GNN-Based Link Prediction via PU-AUC Optimization (https://ieeexplore.ieee.org/document/10869638)

Graph neural networks for link prediction with subgraph sketching (https://arxiv.org/pdf/2209.15486)