Convergence properties of quantum evolutionary algorithms on high dimension problems

Jonathan Wright, Ivan Jordanov

Research output: Contribution to journalArticlepeer-review

210 Downloads (Pure)

Abstract

We propose and investigate new rotation gates for two modified Quantum Inspired Evolutionary methods for solving high dimension optimisation problems. The Quantum Inspired Evolutionary Algorithms (QIEA) were originally used for solving binary encoded problems and their signature features follow superposition of multiple states on a quantum bit and a rotation gate. In order to apply this paradigm to high dimension problems, we propose two quantum methods Half Significant Bit (HSB) and Stepwise Real QEA (SRQEA), developed using binary and real encoding respectively, while keeping close to the original quantum computing metaphor. We introduce five performance metrics and use them to evaluate the proposed approaches against sets of multimodal mathematical test functions and real world problems of high dimensionality. We report issues found while implementing some of the published real QIEA techniques which were the motivation for developing our real algorithm modifications. Our methods focus on introducing and implementing new rotation gate operators used for evolution, including a novel mechanism for preventing premature convergence in the binary algorithm. The applied performance metrics show superior results for our quantum methods on most of the test problems (particularly with high dimension problems), demonstrating faster convergence and accuracy.
Original languageEnglish
Pages (from-to)82-99
Number of pages18
JournalNeurocomputing
Volume326–327
Early online date12 Sep 2017
DOIs
Publication statusPublished - 31 Jan 2019

Keywords

  • quantum evolutionary methods
  • global optimization
  • simulation evaluation
  • estimation of distribution algorithms
  • performance metrics
  • multimodal functions

Fingerprint

Dive into the research topics of 'Convergence properties of quantum evolutionary algorithms on high dimension problems'. Together they form a unique fingerprint.

Cite this