Learning to Sample and Aggregate: Few-shot Reasoning over Temporal Knowledge Graphs

Authors: Ruijie Wang, Zheng Li, Dachun Sun, Shengzhong Liu, Jinning Li, Bing Yin, Tarek Abdelzaher

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, extensive experiments on three real-world TKGs demonstrate the superiority of Meta TKGR over state-of-the-art baselines by a large margin.
Researcher Affiliation Collaboration 1University of Illinois at Urbana Champaign, IL, USA 2Amazon.com Inc, CA, USA
Pseudocode Yes Algorithm 1: Temporal Neighbor Sampler. and Algorithm 2: Meta TKGR: Meta-training.
Open Source Code No The paper does not contain an explicit statement about releasing the source code or a link to a code repository for the methodology described.
Open Datasets Yes Datasets. We evaluate the proposed Meta TKGR framework on three public TKGs, where YAGO [34] and WIKI [22] stores time-varying facts and ICEWS18 [3] is event-centric.
Dataset Splits Yes Given the temporal knowledge graph, we first split the time duration into four with a ratio of 0.4:0.25:0.1:0.25 chronologically, then we collect the entities that firstly appear in each period as background/meta-training/meta-validation/meta-test entity set.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies or library versions (e.g., 'PyTorch 1.9', 'Python 3.8') needed to replicate the experiment.
Experiment Setup Yes For fair comparison, we keep the dimension of all embeddings as 128, and utilize pre-trained 1-shot Trans E embeddings for initialization for models if applicable. We report detailed experimental setup, especially of Meta TKGR, in the Appendix A.4.