Size-efficient sparse population for strictly structured quantum genetic algorithm

Abstract

Quantum genetic algorithm is a field of research to discover a potential structure to realize an effective heuristic, evolutionary optimization technique powered by quantum computation. Apart from contemporary efforts to look for a novel quantum evolutionary design, some studies suggest the idea of strictly structured quantum genetic algorithm, which rigorously imitates the classical practice via a sparse population to sample a few individuals from the given problem's domain, as opposed to the conventional quantum practice that transforms the entire domain into a population itself. Albeit having its own advantages, the algorithm requires to periodically measure and reinitialize the quantum system over generations. It therefore leads to several computational inefficiency issues including the wasteful population initialization, during which individuals that rather marginally contribute to the overall optimization are iteratively generated. This paper proposes that the algorithmic inefficiency from the mentioned problem can be alleviated by preemptively eliminating a large portion of undesirable individuals and still maintain the tasked optimization result to an acceptable degree. The main idea is to deliberately reduce the amount of randomness among the quantum population, thereby discarding individuals that do not contain any fitted genes. A number of tests were conducted on continuous optimization problems to validate the theory of the proposed method, resulting that it can effectively reduce the size of the involved population while minimizing the loss in the original algorithm's performance. © 2022 Elsevier B.V.