Clone Selection Algorithm to Solve Preference Multi-Objective Optimization

Abstract

The difficulty of current multi-objective optimization community lies in the large number of objectives.Lacking enough selection pressure toward the Pareto front, classical algorithms are greatly restrained. In this paper, preference rank immune memory clone selection algorithm (PISA) is proposed to solve the problem of multi-objective optimization with a large number of objectives. The nondominated antibodies are proportionally cloned by their preference ranks, which are defined by their preference information. It is beneficial to increase selection pressure and speed up convergence to the true Pareto-optimal front. Solutions used to approximate the Pareto front can be reduced by preference information. Because only nondominated antibodies are selected to operate, the time complexity of the algorithm can be reduced. Besides, an immune memory population is kept to preserve the nondominated antibodies and ε dominance is employed to maintain the diversity of the immune memory population. Tested in several multi-objective problems with 2 objectives and DTLZ2 and DTLZ3 as high as 8 objectives, PISA is experimentally effective.

A Novel Multiobjective Optimization Method Based on Improved Artificial Bee Colony Algorithm

“…In our experiment, k = 10 is used in DTLZ2, whereas k = 7 is used in DTLZ3. For various algorithm performance evaluation and comparison, we choose generation distance (GD) and spacing (S) as standard [1].…”

“…Because the convergence and distribution of PISA algorithm are better than LbsNSGA-II and NSGA-II [1], we only compare the DMABC algorithm with PISA algorithm. For five objects on DTLZ2 problem, the coverage of PISA algorithm on the first four objects are 0.3 ~ 0.5, on the fifth object is 0.2 ~ 0.5; On the other hand, the coverage of the DMABC algorithm on the all five objects can reach the theoretical value 0 ~ 1.…”

“…So we substitute crowding distance with the harmonic average distance which is calculated by expression (2). 12 1 1 1…”

“…Therefore, multiobjective optimization (MOO) has attracted much attention of researchers and many algorithms have been developed for solving multiobjective optimization problems in the last decade. Preference rank immune memory clone algorithm (PISA) [1], light beam search based multiobjective optimization using evolutionary algorithm (LbsNSGA-II) [2] and non-dominated sorting genetic algorithm II (NSGA-II) [3] are three of the most popular multiobjective algorithms. They have been successfully applied to solve a number of real world multiobjective optimization problems.…”

Multi-Objective Particle Swarm Optimization Algorithm for the Minimum Constraint Removal Problem

“…In phase 2, an elitist group was employed to maintain the current Pareto optimal solution's distribution properties; typical algorithms include the Pareto archived evolutionary strategy (PAES) A by Knowles et al, the strength Pareto evolutionary algorithm (SPEA) A and SPEA2 by Zitzler and Thiele et al [21], and the improved NSGA A (NSGA-II), which is based on an elitist strategy, by Deb and Pratap et al [22]. In phase 3, hybrid algorithms were developed from combinations of the multi-objective evolutionary algorithm and other emerging intelligent algorithms, such as the immune multi-objective optimization algorithm (IMOA) A [23] and multiobjective particle swarm optimization (MOPSO) A [24].…”