FixedSum : a novel algorithm for generating weight vectors in decomposition-based multiobjective optimization

Many multiobjective optimization algorithms employ weight vectors (WVs) for decomposing a problem into multiple subproblems. These weight vectors should be uniformly spread along the Pareto front. Recently, some studies about the development of decomposition-based multiobjective evolutionary algorithms have adopted different methods for generating weight vectors. However, these WVs are often clustered, either near the boundary or in inner regions of the search space. In this paper, we have proposed a novel algorithm, FixedSum, for the generation of an arbitrary number of WVs of any user-specified dimension. These weight vectors are more uniformly spread in the search space, and we have compared our results with other methods on 5D, 8D, 10D and 12D weight vectors. For further validation, we have applied our weight vectors along with the WVs of two other methods for solving the DTLZ problems. All the results demonstrate the improved spreadability of our method compared to competing approaches.

---

Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki (2025).