|
Frontiers of Information Technology & Electronic Engineering
ISSN 2095-9184 (print), ISSN 2095-9230 (online)
2019 Vol.20 No.10 P.1344-1360
Unusual phenomenon of optimizing the Griewank function with the increase of dimension
Abstract: The Griewank function is a typical multimodal benchmark function, composed of a quadratic convex function and an oscillatory nonconvex function. The comparative importance of Griewank’s two major parts alters in different dimensions. Different from most test functions, an unusual phenomenon appears when optimizing the Griewank function. The Griewank function first becomes more difficult and then becomes easier to optimize with the increase of dimension. In this study, from the methodology perspective, this phenomenon is explained by structural, mathematical, and quantum analyses. Furthermore, frequency transformation and amplitude transformation are implemented on the Griewank function to make a generalization. The multi-scale quantum harmonic oscillator algorithm (MQHOA) with quantum tunnel effect is used to verify its characteristics. Experimental results indicate that the Griewank function’s two-scale structure is the main reason for this phenomenon. The quantum tunneling mechanism mentioned in this paper is an effective method which can be generalized to analyze the generation and variation of solutions for numerous swarm optimization algorithms.
Key words: Griewank, Two-scale structure, Multi-scale quantum harmonic oscillator algorithm, Quantum tunnel effect
关键词组:
References:
Open peer comments: Debate/Discuss/Question/Opinion
<1>
DOI:
10.1631/FITEE.1900155
CLC number:
TP3-05
Download Full Text:
Downloaded:
2105
Download summary:
<Click Here>Downloaded:
1760Clicked:
5141
Cited:
0
On-line Access:
2019-11-11
Received:
2019-03-19
Revision Accepted:
2019-08-21
Crosschecked:
2019-10-10