配方控制解码器审计用于结构知识图补全

Recipe-Controlled Decoder Audit for Structural Knowledge-Graph Completion

精选理由

搞清解码器选择的关键因素

AI 摘要

本文提出配方控制的解码器审计(RCDA)用于结构知识图补全。以ComplEx和DistMult为主控对,辅以RotatE和TransE点检,在7个基准上评估。5个标准KG上ComplEx与DistMult的MRR差异在+0.005至+0.012之间。小KG上解码器效应更显著:Kinship中ComplEx优势达+0.143 MRR(6种子),UMLS中优势为+0.022 MRR(6种子)。YAGO3-10上,该配方下L=0的ComplEx在d=128时达到0.6971±0.0048 MRR。

AI 翻译 · 中文

本文提出配方控制的解码器审计(RCDA)用于结构知识图补全。以ComplEx和DistMult为主控对,辅以RotatE和TransE点检,在7个基准上评估。5个标准KG上ComplEx与DistMult的MRR差异在+0.005至+0.012之间。小KG上解码器效应更显著:Kinship中ComplEx优势达+0.143 MRR(6种子),UMLS中优势为+0.022 MRR(6种子)。YAGO3-10上,该配方下L=0的ComplEx在d=128时达到0.6971±0.0048 MRR。

arXiv cs.LGWe present a recipe-controlled decoder audit (RCDA) for structural transductive knowledge-graph completion (KGC). The audit asks a simple reporting question: before attributing gains to an encoder or training recipe, wha