Details

Autor(en) / Beteiligte

• Tolga Altinoz, O.
• Deb, Kalyanmoy
• Egemen Yilmaz, A.

Titel

Distributed approaches for reference-point-based multi-objective hybrid problems

Ist Teil von

• Information sciences, 2018-10, Vol.467, p.323-341

Ort / Verlag

Elsevier Inc

Erscheinungsjahr

2018

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• •Hybrid multi-objective optimization problems are proposed for distributed computing.•A set of hybrid benchmark problems are defined.•Boundary violation modules for algorithms named as HHA, FFHHA, and CA are introduced to increase the performance.•The implementations for hybrid problems are grouped and discussed on SPSP & MPSP. Testing and performance comparisons for optimization algorithms and methods are an important part of demonstrating accurate behavior. These tests are accomplished using numerical and graphical illustrations of the results obtained from the proposed algorithms. To emphasize the advantages and disadvantages of the proposed approaches and algorithms, a set of problems for which the solution and common properties are known is needed. Thus, the behavior of the algorithm as it obtains the solution set can be explained by using the common properties of the test problems. Therefore, a set of well-known benchmark problems has been proposed by researchers, and a portion of these problems is specifically designed for testing multi-objective optimization algorithms. Although these problems are sufficient to present the performances of optimization algorithms, there is no problem set for investigating the distributed performance of optimization algorithms. Hence, a method for the performance comparison of distribution methods for multi-objective optimization algorithms is needed. In this study, a set of new test problems, called hybrid problems, is defined by aligning two different well-known test functions for parallelization models. These novel problems are solved using the distributed models. Lastly, a set of approaches is proposed to increase the performance of any similar distributed models.