TorusE: Knowledge Graph Embedding on a Lie Group
Authors: Takuma Ebisu, Ryutaro Ichise
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experiments We evaluated Torus E from two perspectives: one is its scalability and the other is the accuracies of the link prediction tasks. ... 5.3 Results The results of the link prediction tasks are shown in Table 3. |
| Researcher Affiliation | Academia | Takuma Ebisu,1,2 Ryutaro Ichise2,1,3 1SOKENDAI (The Graduate University for Advanced Studies) 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo, Japan 2National Institute of Informatics 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo, Japan 3National Institute of Advamced Industrial Science and Technology 2-3-26 Aomi, Koto-ku, Tokyo, Japan {takuma,ichise}@nii.ac.jp |
| Pseudocode | No | The paper describes methods in prose and mathematical formulas, but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any explicit statement or link for open-source code availability for the described methodology. |
| Open Datasets | Yes | The experiments are conducted on two benchmark datasets: WN18 and FB15K (Bordes et al. 2013). |
| Dataset Splits | Yes | The details of the datasets are shown in Table 2. ... The best models were selected by the MRR with filtered rankings on the validation set. |
| Hardware Specification | Yes | They are measured by using a single GPU (NVIDIA Titan X). |
| Software Dependencies | No | The paper mentions implementation details (e.g., stochastic gradient descent, Bern method for negative sampling) but does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, etc., with their versions). |
| Experiment Setup | Yes | We conducted a grid search to find suitable hyperparameters for each dataset. The dimension was fixed to 10000, because a model with a higher dimension yields a better result in practice. We selected the margin γ from {2000, 1000, 500, 200, 100} and the learning rate α from {0.002, 0.001, 0.0005, 0.0002, 0.0001}. Scoring functions fd were selected from { f L1, f L2, fe L2}. The optimal configurations were as follows: γ = 2000, α = 0.0005 and fd = f L1 for WN18; γ = 500, α = 0.001 and fd = fe L2 for FB15K. |