SqueezeFit: Label-Aware Dimensionality Reduction by Semidefinite Programming

DOI10.1109/TIT.2019.2962681MaRDI QIDQ5124405zbMATH OpenOpenAlexFDO

Authors Culver McWhirter, Dustin G. Mixon, Soledad Villar

Publication date 29 September 2020

Published in IEEE Transactions on Information Theory (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1812.02768

Mathematics Subject Classification ID

Semidefinite programming (90C22) Sampling theory in information and communication theory (94A20)

Abstract: Given labeled points in a high-dimensional vector space, we seek a low-dimensional subspace such that projecting onto this subspace maintains some prescribed distance between points of differing labels. Intended applications include compressive classification. Taking inspiration from large margin nearest neighbor classification, this paper introduces a semidefinite relaxation of this problem. Unlike its predecessors, this relaxation is amenable to theoretical analysis, allowing us to provably recover a planted projection operator from the data.

Recommendations

• Supervised dimensionality reduction via sequential semidefinite programming
• A unified framework for semi-supervised dimensionality reduction
• Dimension reduction with semi-supervised pairwise covariance-preserving projection
• A general soft label based linear discriminant analysis for semi-supervised dimensionality reduction
• Sufficient dimension reduction via principal L\(q\) support vector machine
• Constrained large margin local projection algorithms and extensions for multimodal dimensionality reduction
• Minimal \(\sigma\)-field for flexible sufficient dimension reduction

Cited in

(3)

• Lie PCA: density estimation for symmetric manifolds
• Estimation of Wasserstein distances in the spiked transport model
• Matching component analysis for transfer learning

This page was built for publication: SqueezeFit: Label-Aware Dimensionality Reduction by Semidefinite Programming

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5124405)