Harvesting Preconditioning for Scalable High-Resolution Signal and Image Reconstruction

Abstract

The goal of the proposed research program is to accelerate ill-conditioned signal and image reconstruction via the power of preconditioning, using a novel trajectory-based analysis framework that tightly integrates the statistical and optimization aspects of high-dimensional signal estimation in an intertwined manner. The proposed research program, if successful, will establish a comprehensive framework of understanding and designing tailored first-order methods with accelerated convergence rates for ill-conditioned nonconvex signal and image reconstruction with provable statistical, computational and robustness guarantees, which are of critical importance for extracting fine-grained and weak information in resource-constrained and sample-starved environments.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 06, 2025
Source ID
FA95502510060

Entities

People

  • Yuejie Chi

Organizations

  • Air Force Office of Scientific Research
  • Carnegie Mellon University
  • United States Air Force

Tags

Readers

  • Distributed Systems and Data Platform Development
  • Image Processing and Computer Vision.
  • Linear Algebra