Interstellar: Searching Recurrent Architecture for Knowledge Graph Embedding

Authors: Yongqi Zhang, Quanming Yao, Lei Chen

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on real datasets demonstrate the effectiveness of the searched models and the efficiency of the proposed hybrid-search algorithm.
Researcher Affiliation Collaboration Yongqi Zhang1,3 Quanming Yao1,2 Lei Chen3 14Paradigm Inc. 2Department of Electronic Engineering, Tsinghua University 3Department of Computer Science and Engineering, HKUST
Pseudocode Yes Algorithm 1 Proposed search recurrent architecture as the Interstellar algorithm.
Open Source Code Yes 1Code is available at https://github.com/Auto ML-4Paradigm/Interstellar, and correspondence is to Q. Yao.
Open Datasets Yes Here, we illustrate the designed search space A in Section 3.1 using Countries [8] dataset... We use four cross-lingual and cross-database subset from DBpedia and Wikidata generated by [18]... We use three famous benchmark datasets, WN18-RR [13] and FB15k-237 [48], which are more realistic than their superset WN18 and FB15k [7], and YAGO3-10 [31], a much larger dataset.
Dataset Splits No The paper mentions using 'validation set' but does not explicitly provide specific training/validation/test dataset splits (e.g., percentages or sample counts) within the main text. It defers some details to external papers (e.g., '[18]') or appendix without providing the concrete numbers.
Hardware Specification Yes Experiments are written in Python with Py Torch framework [35] and run on a single 2080Ti GPU.
Software Dependencies No The paper mentions 'Python with Py Torch framework [35]' but does not provide specific version numbers for Python, PyTorch, or other software dependencies.
Experiment Setup No The paper states 'Training details of each task are given in Appendix A.5.' and mentions 'hyper-parameters, i.e. learning rate, decay rate, dropout rate, L2 penalty and batch-size (details in Appendix A.5)', but the specific values for these experimental setup details are not provided in the main text.