ML ICML

Principal Component Hierarchy for Sparse Quadratic Programs

May 24, 2021

We propose a novel approximation hierarchy for cardinality-constrained, convex quadratic programs that exploits the rank-dominating eigenvectors of the quadratic matrix. Each level of approximation admits a min-max characterization whose objective function can be optimized over the binary variables analytically, while preserving convexity in the continuous variables. Exploiting this property, we propose two scalable optimization algorithms, coined as the “best response” and the “dual program”, that can efficiently screen the potential indices of the nonzero elements of the original program. We show that the proposed methods are competitive with the existing screening methods in the current sparse regression literature, and it is particularly fast on instances with high number of measurements in experiments with both synthetic and real datasets.

Overall

< 1 minute

Robbie Vreugdenhil, Viet Anh Nguyen, Armin Eftekhari, Peyman Mohajerin Esfahani

ICML 2021

Share Article

Related publications

ML ICLR Top Tier
February 19, 2024

Nguyen Hung-Quang, Yingjie Lao, Tung Pham, Kok-Seng Wong, Khoa D Doan

CV ML AAAI Top Tier
January 8, 2024

Tran Huynh Ngoc, Dang Minh Nguyen, Tung Pham, Anh Tran

ML AAAI Top Tier
January 8, 2024

Viet Nguyen*, Giang Vu*, Tung Nguyen Thanh, Khoat Than, Toan Tran

ML NeurIPS Top Tier
October 4, 2023

Van-Anh Nguyen, Trung Le, Anh Tuan Bui, Thanh-Toan Do, Dinh Phung