Nonconvex Information Processing for Heterogeneous and Distributed Data

Abstract

The objective of this research program is to develop a comprehensive framework for nonconvex information processing with large-scale heterogeneous and distributed data. We aim to design efficient, robust and provably accurate algorithms for information processing using their natural nonconvex formulations without resorting to expensive convex relaxations. We will pursue a deeper understanding of geometric properties of nonconvex loss surfaces and optimization trajectories, with the emphasis on more realistic data models that can be heterogeneous and distributed. Novel combinations of insights and techniques from statistical signal processing, mathematical optimization, information theory, and high-dimensional statistics will be developed throughout the proposed research program to meet the research objectives

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 28, 2021
Accession Number
AD1197155

Entities

People

  • Yuxin Chen

Organizations

  • Princeton University

Tags

DTIC Thesaurus Topics

  • Abstracts
  • Algorithms
  • Convex Programming
  • Data Science
  • Engineering
  • Estimators
  • Geometry
  • Information Processing
  • Information Science
  • Information Systems
  • Information Theory
  • Machine Learning
  • Mathematical Programming
  • Mathematics
  • Measurement
  • Neural Networks
  • Operations Research
  • Optimization
  • Signal Processing
  • Statistical Estimation
  • Statistics

Readers

  • Distributed Systems and Data Platform Development
  • Operations Research