A Survey on Representation Learning for Graph Reasoning

PhD Qualifying Examination


Title: "A Survey on Representation Learning for Graph Reasoning"

by

Mr. Xin LIU


Abstract:

Embedding graphs into continuous vector spaces is a focus of current 
research. Mining valuable hidden information from such spaces relies on 
the support of reasoning technology. Representation learning for 
homogeneous graphs mainly aims to capture the information of topology and 
geometry. Reasoning over heterogeneous graphs is more challenging. 
Inspired by distributed embeddings for homogeneous graphs, the derivation 
of basic operations and neural networks for heterogeneous is also under 
development. Knowledge graphs, providing well-structured relational 
information between entities, have gained much attention in artificial 
intelligence. Reasoning over them is also important because it can not 
only infer new facts from existing data but provide interpretations for 
downstream tasks. In this survey, we introduce the basic concept and 
definitions of graph reasoning and the research work for embedding methods 
for reasoning over graphs. Specifically, we dissect the reasoning methods 
into two categories: shallow representation-based, and graph 
convolution-based reasoning. Finally, we summarize real-world applications 
of graph reasoning, such as knowledge graph completion, question 
answering, semantic segmentation, and so on.


Date:			Tuesday, 11 August 2020

Time:                  	2:00pm - 4:00pm

Zoom Meeting:		https://hkust.zoom.us/j/99100964120

Committee Members:	Dr. Yangqiu Song (Supervisor)
   			Prof. Nevin Zhang (Chairperson)
  			Dr. Qiong Luo
  			Prof. Dit-Yan Yeung


**** ALL are Welcome ****