On Link Prediction in Knowledge Bases: Max-K Criterion and Prediction
Protocols
PRESENTER:
Yongyi Mao
University of Ottawa
ABSTRACT:
Building knowledge base embedding models for link prediction has achieved
great success. We however argue that the conventional top-$k$ criterion
used for evaluating the model performance is inappropriate. This work
introduces a new criterion, referred to as max-$k$. Through theoretical
analysis and experimental study, we show that the
top-$k$ criterion is fundamentally inferior to max-$k$. We also introduce
two prediction protocols for the max-$k$ criterion. These protocols are
strongly justified theoretically. Various insights concerning the max-$k$
criterion and the two protocols are obtained through extensive
experiments. This is a joint work with Jiajie Mei, Richong Zhang and Ting
Deng at Beihang University, China. The work has been accepted to SIGIR
2018.
Bio: Yongyi Mao received his Bachelor of Engineering degree at the
Southeast University (Nanjing, China) in 1992. In 1995, he received his
medical degree at Nanjing Medical University (Nanjing, China). In 1998,
Yongyi Mao obtained his Master of Science degree at the University of
Toronto, in the Department of Medical Biophysics . In 2003, he completed
his PhD in electrical engineering at the University of Toronto and joined
the faculty of School of EECS at the University of Ottawa as an Assistant
Professor. He was promoted to Associate Professor in 2008 and then to Full
Professor in 2012. Yongyi Mao's research interest at this moment includes
machine learning, information theory and their applications in knowledge
bases and natural language processing.