Heterogeneous representation learning and matching for few-shot relation prediction
作者：
Highlights：
• A novel few-shot relation prediction method is proposed to capture the heterogeneous influences of relational neighbors and their features.
• Single-layer CNN with differently sized filters is devised to capture multi-scale characteristics while controlling model complexity.
• Multiple similarity methods are taken into consideration to construct a matching metric between query and support set.
摘要
•A novel few-shot relation prediction method is proposed to capture the heterogeneous influences of relational neighbors and their features.•Single-layer CNN with differently sized filters is devised to capture multi-scale characteristics while controlling model complexity.•Multiple similarity methods are taken into consideration to construct a matching metric between query and support set.
论文关键词：Knowledge graphs,Few-shot learning,Relation prediction,Representation learning,Convolutional network
论文评审过程：Received 9 January 2022, Revised 25 May 2022, Accepted 3 June 2022, Available online 9 June 2022, Version of Record 14 June 2022.
论文官网地址：https://doi.org/10.1016/j.patcog.2022.108830
