Decomposing a Multiobjective Optimization Problem into a Number of Reduced-Dimension Multiobjective Subproblems Using Tomographic Scanning

In this paper, we design a novel method to handle multi-and many-objective optimization problem. The proposed method adopts the idea of tomographic scanning in medical imaging to decompose the objective space into a combination of many tomographic maps to reduce the dimension of objectives incrementally. Moreover, subpopulations belonging to different tomographic maps can help each other in evolving the optimal results. We compared the performance of the proposed algorithm with some classical algorithms such as NSGA-II and MOEA/DTCH and their state-of-the-art variants including MOEA/DDE, NSGA-III and MOEA/D-PBI. The experimental results demonstrate that the proposed method significantly outperforms MOEA/D-TCH, MOEA/D-DE and NSGA-II, and is very competitive with MOEA/D-PBI and NSGA-III in terms of convergence speed.