New Method Focused on Data Mining from Multi-Relational Data Accepted to IJCAI 2018

Wed Jun 20, 2018

Kai Liu has recently had his paper, “High-Order Co-Clustering via Strictly Orthogonal and Symmetric L1-Norm Nonnegative Matrix Tri-Factorization,” accepted into the 27th International Joint Conference on Artificial Intelligence (IJCAI 2018).

This work focuses on data mining from multi-relational data and derives the solution by making use of Alternating Direction Method of Multipliers (ADMM). Different from classical Mulplicative Updating Algorithm (MUA), the proposed method can get the solution which is strictly orthogonal to avoid soft-clustering. Extensive experiments illustrate its advantages including less time consumption, and lower objective value. Moreover, the paper shows a new approach for L-1 norm objective optimization, which is easily extended to squared Frobenius norm, L2,1 norm or any other differentiable or non-differentiable objective.

This paper will be presented this July in Stockholm, Sweden. IJCAI is the top conference in Artificial Intelligence.