Top Paper of Korea Data Mining Academy
Top Paper of Korea Data Mining Academy
  • Reporter Park Do-won
  • 승인 2014.03.05 16:50
  • 댓글 0
이 기사를 공유합니다

CSE graduate student, Yongdeok Kim (M.S.-Ph.D. integrated, Advisor Professor Seungjin Choi) won the best paper award of SAS paper award contest, at the autumn symposium of Korea Business Intelligence Data Mining Society (KDMS). The awarded paper is “Scalable Variational Bayeisan Matrix Factorization”.
In the paper, he presents a method for improving matrix factorization technique. Although Bayesian treatment for matrix factorization can effectively lighten the over fitting problem, it has limitations in that existing inference algorithms for Bayesian matrix factorization are fit for web-scale datasets with numerous users and ratings. To solve this, Kim proposed a scalable learning algorithm for variational Bayesian matrix factorization.
KDMS is an academic society that was established in 2001 to develop, propagate, and apply techniques related to data mining, and to promote the development of national information technology. This autumn symposium was held on Nov. 29-30 at Sokcho Delpino Resort.