Unsupervised/Supervised Column Subset Selection
- The main idea is to apply the combinatorial search on the column subset graph. Heuristic functions based on eigenvalues, and efficient implementation methods were introduced.The first non-trivial optimal algorithm for supervised column subset selection for multi-target prediction was proposed.
- Unsupervised version was published by AAAI 2019; Supervised version was published by IJCAI 2021.
Outlier Detection for Principal Component Analysis
- The main idea is to formulate the optimization task as a graph search problem and use combinatorial search to solve it. The centered rank-one modification for eigenvalue decomposition was proposed for efficient implementation.
- Papers were published by AAAI 2021, ICDM2021.