A multi-objective evolutionary algorithm with decomposition and the information feedback for high-dimensional medical data

Wang, Mingjing; Heidari, Ali Asghar; Chen, Huiling

doi:10.1016/j.asoc.2023.110102

Cited by 9 publications

(1 citation statement)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is achieved by converting a multi-objective problem into multiple single-objective sub-problems that can be solved individually, and utilizing a neighborhood search strategy and weight vectors to ensure global convergence and diversity. Compared to other multi-objective optimization algorithms, such as the dominance-based multi-objective optimization algorithm (NSGA-II), MOEA/D has several advantages: the MOEA/D algorithm uses a decomposition strategy for solving, so it can effectively deal with high-dimensional problems [9][10][11]; approximation of Pareto optimal solutions by collaborative solving among subproblems [12,13]; and using a single reference point to guide solution generation reduces the number of depth evaluations and improves the efficiency of the algorithm by only evaluating the solution in the vicinity of the reference point in each problem [14][15][16].…”

Section: Introductionmentioning

confidence: 99%

A Dynamic Parameter Tuning Strategy for Decomposition-Based Multi-Objective Evolutionary Algorithms

Zheng,

Ning,

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

The penalty-based boundary cross-aggregation (PBI) method is a common decomposition method of the MOEA/D algorithm, but the strategy of using a fixed penalty parameter in the boundary cross-aggregation function affects the convergence of the populations to a certain extent and is not conducive to the maintenance of the diversity of boundary solutions. To address the above problems, this paper proposes a penalty boundary crossing strategy (DPA) for MOEA/D to adaptively adjust the penalty parameter. The strategy adjusts the penalty parameter values according to the state of uniform distribution of solutions around the weight vectors in the current iteration period, thus helping the optimization process to balance convergence and diversity. In the experimental part, we tested the MOEA/D-DPA algorithm with several MOEA/D improved algorithms on the classical test set. The results show that the MOEA/D with the DPA has better performance than the MOEA/D with the other decomposition strategies.

show abstract