Structured Embedding via Pairwise Relations and Long-Range Interactions in Knowledge Base
Authors: Fei Wu, Jun Song, Yi Yang, Xi Li, Zhongfei Zhang, Yueting Zhuang
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments and Results We compare our Se PLi with several state-of-the-art knowledge embedding methods as follows Four benchmark datasets are used for the performance evaluation. Table 3 and Table 4 shows the results of each compared model over Kinship and UMLS. |
| Researcher Affiliation | Academia | Fei Wu and Jun Song College of Computer Science Zhejiang University, China Yi Yang Centre for Quantum Computation and Intelligent Systems University of Technology, Sydney Xi Li College of Computer Science Zhejiang University, China Zhongfei Zhang Department of Information Science and Electronic Engineering Zhejiang University, China Yueting Zhuang College of Computer Science Zhejiang University, China |
| Pseudocode | Yes | Algorithm 1 summarizes the procedure of our proposed Se PLi . |
| Open Source Code | No | The paper does not provide any statement or link about the availability of open-source code for the described methodology. |
| Open Datasets | Yes | Four benchmark datasets are used for the performance evaluation. We divide each dataset into three distinct parts: observed, unobserved entities, and unobserved relations. Kinship Alyawarra Kinship data (Denham 1973) records the family relations of Alyawarra... UMLS Unified Medical Language System (UMLS) is an upper-level ontology dataset created by Mc Cray (Mc Cray 2003)... Word Net (Miller 1995) is an online lexical dataset... Freebase Following (Socher et al. 2013), we obtained a subset of Freebase from the People domain. |
| Dataset Splits | No | The paper mentions 'No. of observed facts(train/test)' in Table 1, indicating train and test splits, and states 'hyper-parameters are set using grid search'. However, it does not explicitly provide details about a distinct validation set split or its use for hyperparameter tuning separate from the test set. |
| Hardware Specification | Yes | This experiment is conducted on a Intel Core i7-4790 CPU @ 3.6GHz. |
| Software Dependencies | No | The paper mentions 'minibatched Limited-memory BFGS (L-BFGS)' as an optimization algorithm, but does not provide specific version numbers for any software dependencies (e.g., programming languages, libraries, or frameworks). |
| Experiment Setup | Yes | Parameter settings In the experiments, hyper-parameters are set using grid search: (i) entity vectors, relation matrixes, and biases are all initialized by a uniform distribution on [ 0.001, 0.001]; (ii) for datasets Kinship (Denham 1973) and UMLS (Mc Cray 2003), the dimensionality of the embedding space is set to d = 10; for datasets Word Net (Miller 1995) and Free Base (Bollacker et al. 2008), the dimensionality of the embedding space is set to d = 100; (iii) the weighting parameter of long-range loss is λ1 = 0.01, and the regularization parameter is λ2 = 0.0001; (iv) the length of each path is no longer than 4 (L = 4) and the maximum number of long-range interactions for each pairwise relation is K = 10; (v) minibatched Limited-memory BFGS (L-BFGS) (Malouf 2002; Andrew and Gao 2007) is used for optimizing non-convex object function and the number of training iterations is set to 500. |