Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4961479 | Procedia Computer Science | 2017 | 8 Pages |
Abstract
This paper presents a method for permanent parameter tuning of optimization algorithms. We describe mathematical formalization of parameter tuning problem and meta-model used for it. There are also the scheme of the proposed method, the description of search stages and meta-algorithms used. A way to classify optimization problems is also introduced. It defines how the tuning process will be divided for different types of problems in order to reuse tuning results. Besides, several ways to organize tuning process are suggested. The proposed method is implemented in the program system for automated parameter tuning. The results of experiments indicate the effectiveness of developed system.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Science (General)
Authors
T. Agasiev, A. Karpenko,