In the new semester, our Lab, Web Search Mining and Security Lab, plans to hold an academic seminar every Monday, and each time a keynote speaker will share understanding of papers on his\/her related research with you.
<\/p>\n\n\n\n
This week’s seminar is organized as follows: <\/p>\n\n\n\n
Everyone interested in this topic is welcomed to join us. the following is the abstract for this week\u2019s paper.<\/p>\n\n\n\n
\n\t
\n\t\t\tVariational Knowledge Graph Reasoning\n\t\t<\/p>\n\t\t
\n\t\t\tWenhu Chen, Wenhan\nXiong, Xifeng Yan, William Yang Wang\n\t\t<\/p>\n\t\t
\n\t\t\tAbstract\n\t\t<\/p>\n\t\t
\n\t\t\t Inferring missing links in knowledge\ngraphs (KG) has attracted a lot of attention from the research community. In\nthis paper, we tackle a practical query answering task involving predicting the\nrelation of a given entity pair. We frame this prediction problem as an\ninference problem in a probabilistic graphical model and aim at resolving it\nfrom a variational inference perspective. In order to model the relation\nbetween the query entity pair, we assume that there exists an underlying latent\nvariable (paths connecting two nodes) in the KG, which carries the equivalent\nsemantics of their relations.<\/span>\n\t\t<\/p>\n\t\t \n\t\t\t However, due to the intractability of\nconnections in large KGs, we propose to use variation inference to maximize the\nevidence lower bound. More specifically, our framework (DIVA) is composed of\nthree modules, i.e. a posterior approximator, a prior (path finder), and a\nlikelihood (path reasoner). By using variational inference, we are able to\nincorporate them closely into a unified architecture and jointly optimize them\nto perform KG reasoning. With active interactions among these sub-modules, DIVA\nis better at handling noise and coping with more complex reasoning scenarios.\nIn order to evaluate our method, we conduct the experiment of the link\nprediction task on multiple datasets and achieve state-of-the-art performances\non both datasets.<\/span>\n\t\t<\/p>\n\t\t \n\t\t\t<\/span>\n\t\t<\/p>\n\t<\/div>\n<\/p>\n\n\n\n Last Monday, Qinghong Jiang<\/strong> gave a presentation about the paper, Variational Knowledge Graph Reasoning<\/strong>, and shared some opinion on it.<\/p>\n\n\n\n This paper proposed to combine these two steps – \u201cPath-Finding\u201d and \u201cPath- The authors explained why is the results on the NELL dataset much smaller than results on the FB15k dataset, because NELL is a simple dataset, but FB15k is much harder than NELL and arguably more relevant for real-world scenarios. <\/p>\n","protected":false},"excerpt":{"rendered":" NLPIR SEMINAR Y2019#6 INTRO In the new s … \u7ee7\u7eed\u9605\u8bfb NLPIR SEMINAR 19th ISSUE COMPLETED<\/strong><\/h2>\n\n\n\n
Reasoning\u201d – together as a whole from the perspective of the latent variable graphic model. This graphic model views the paths as discrete latent variables and relation as the observed variables with a given entity pair as the condition, thus the path-finding module can be viewed as a prior distribution to infer the underlying links in the KG. In contrast, the path reasoning module can be viewed as the likelihood distribution, which classifies underlying links into multiple classes.<\/p>\n\n\n\n