Optimizing Gradient Methods for IoT Applications

Eghbal Hosseini, Line Reinhardt, Danda B. Rawat

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

Abstract

Solving Linear Programming (LP) and Non-Linear Programming (NLP) problems are momentous because of their wide applications in real-life problems. There is no unified way to find the global optimum for NLPs. But on the other hand, Simplex Algorithm, as the dominating methodology for LPs for several decades moves only on the boundary (vertices) and ignores the vast majority of the feasible region in the process of searching. In this paper, we study two gradient-based methodologies that explore the whole feasible region, which guarantee faster convergence rates for both LP and NLP optimization problems including IoT problems such as Software Defined Internet of Vehicles (SDIoV) and Vehicular Ad hoc Networks (VANETs). The Gradient-Simplex Algorithm (GSA) for LPs, which moves inside the feasible region in the gradient direction at first to reduce the search space and then explores the reduced boundary to find an optimal solution. The Evolutionary-Gradient Algorithm (EGA), on the other hand, is for NLPs and uses an evolutionary population to estimate gradients by evolving to find better solutions in steps. Based on extensive simulations, obtained numerical results show that both approaches provide efficient solutions and outperform the state-of-the-art methods on optimization problems with large feasible spaces. Comparative results of applying the GSA on SDIoV and VANETs with different sizes are included.

OriginalsprogEngelsk
TidsskriftIEEE Internet of Things Journal
Vol/bind9
Udgave nummer15
Sider (fra-til)13694-13704
Antal sider11
ISSN2327-4662
DOI
StatusUdgivet - 1 aug. 2022

Bibliografisk note

Important note from the Publisher regarding the attached version of the article: “© 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.”

Emneord

  • Approximation algorithms
  • Evolutionary Population
  • Feasible Region
  • Gradient Function
  • Internet of Things
  • Linear programming
  • Optimization
  • Simplex Algorithm
  • Sociology
  • Software algorithms
  • Statistics

Citer dette