《顶级会议论坛PPT:何高乐.pptx》由会员分享,可在线阅读,更多相关《顶级会议论坛PPT:何高乐.pptx(38页珍藏版)》请在三个皮匠报告上搜索。
1、Mining Implicit Entity Preference from User-Item Interaction Data for Knowledge Graph Completion via Adversarial Learning,Gaole He, Junyi Li, Wayne Xin Zhao, Peiju Liu and Ji-Rong Wen,Outline,Introduction Background Motivation Related Work Task Definition Challenge , ) evaluates whether the entity c
2、an be the answer to a given query , , ? by computing the following probability,MLP-1,MLP-2,KG,score,query,answer,Query-specific Entity Generator,For each query , , ?, the probability distribution to sample a candidate entity from candidate entity set , is defined as,Training,For discriminator, minim
3、ize the following loss over the real answer and the fake samples from the generator,For the generator, sample fake samples from the candidate entities, and then train it with policy gradient,Outline,Introduction Challenge & Solution Experiment Datasets Setup Results & Analysis Case Study Conclusion,Datasets,Item-entity alignment from KB4Rec dataset Three domains: movie, music and book Training: va