EPICS: error-preserved and interpolation-corrected surrogate-assisted particle swarm optimization for complex optimization
Abstract Surrogate-assisted particle swarm optimization (SAPSO) has been proven to be efficient in solving high-dimension problems. However, the error originating from the surrogate model tends to mislead search direction and frequently traps in a local optimum. Therefore, this paper proposes error-...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-06-01
|
| Series: | Complex & Intelligent Systems |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s40747-025-01947-0 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Surrogate-assisted particle swarm optimization (SAPSO) has been proven to be efficient in solving high-dimension problems. However, the error originating from the surrogate model tends to mislead search direction and frequently traps in a local optimum. Therefore, this paper proposes error-preserved and interpolation-corrected SAPSO (EPICS) for complex optimization. Firstly, the error between estimated evaluation and real evaluation acquires revision through preserved error to better represent practical particles. Then, the interpolation-corrected mechanism is embedded between particles and the surrogate model to correct particles resisting premature convergence. Finally, experimental results on CEC2013 benchmark functions denote EPICS has advantages in more accurate fitness evaluation for high-dimension problems, which suggests that the search direction obtains proofreading, and the exploitation ability achieves enhancement with the correction of interpolation. Additionally, well-known complex functions are tested to validate the effectiveness of EPICS compared to state-of-the-art algorithms. Simulations indicate EPICS presents a promising searching ability for complex optimization, especially for high-dimensions. Finally, we applied EPICS to a complex truss stress optimization problem, resulting in lower stress and a more uniform stress distribution. |
|---|---|
| ISSN: | 2199-4536 2198-6053 |