Max Berrendorf
Max Berrendorf
This PR updates the README from a model-centric to a more (representation + interaction)-centric view
This PR adds a command to compute post-adjusted metrics for all reproducibility configurations shipped with PyKEEN.
This PR adds a new metric to the rank-based metric results, which measures the mean difference between (relative) optimistic and pessimistic ranks. Thus, a positive value indicates that there are...
This draft PR aims at splitting datasets from their loading procedure: * one dataset class * multiple different data loaders
### Publication Link https://arxiv.org/abs/2202.07919 ### Reference Implementation https://github.com/anrep/HousE ### Additional Implementations _No response_ ### Relevance [accepted](https://icml.cc/Conferences/2022/Schedule?showEvent=17408) at ICML’22 ### Additional Context The model is defined [here](https://github.com/anrep/HousE/blob/main/codes/model.py). In paper the discuss...
**Is your feature request related to a problem? Please describe.** While we can get variances in (rank-based) metrics by repeated training of a model, there is a cheaper alternative which...
**Is your feature request related to a problem? Please describe.** The inverse triples usage is currently somewhat inconsistent, e.g., we expected the training instances to contain inverse triples, but do...
**On which paper(s) is your requested model based.** * Conference Proceedings: https://ojs.aaai.org//index.php/AAAI/article/view/5694 * Preprint: https://arxiv.org/abs/1911.00219 * Supplementary: https://shikhar-vashishth.github.io/assets/pdf/interacte_supp.pdf **List the original and other existing implementations of this model** * official...
**Describe the benchmark KG dataset**: A subset of http://rtw.ml.cmu.edu/rtw/ Besides normal triples, there are also explicit positive **and negative** labels for the triple classification task. **Dataset is pre-stratified into train/test/valid**...
**Is your feature request related to a problem? Please describe.** [torch.hub](https://pytorch.org/docs/stable/hub.html) allows to share pre-trained models. Providing pre-trained models enables re-use for down-stream applications without the need to train your...