Preference-inspired coevolutionary algorithm with active diversity strategy for multi-objective multi-modal optimization

Rui Wang,Wubin Ma,Mao Tan,Guohua Wu,Ling Wang,Dunwei Gong,Jian Xiong

DOI: https://doi.org/10.1016/j.ins.2020.09.075

IF: 8.1

2021-02-01

Information Sciences

Abstract:<p>Multi-objective multi-modal optimization problems have recently received increasing attention in the field of evolutionary computation. Addressing such problems is not easy for existing evolutionary multi-objective algorithms (EMOAs) since they require finding solutions with good convergence and diversity in both objective and decision spaces. This study therefore proposes a new algorithm, namely, the preference-inspired coevolutionary algorithm (PICEAg) with an active diversity strategy, to deal with multi-objective multi-modal optimization problems. The proposed algorithm, denoted as MMPICEAg, adopts the popular coevolutionary framework of PICEAg and introduces a diversity-aware fitness assignment and a double-diversity archive update strategy to promote diversity in objective and decision spaces simultaneously. The performance of MMPICEAg is compared with that of three general EMOAs as well as four state-of-the-art multi-modal EMOAs. The comparison results on three sets of widely used benchmarks clearly demonstrate the effectiveness of MMPICEAg for multi-objective multi-modal optimization.</p>

computer science, information systems

What problem does this paper attempt to address?

This paper attempts to address the problem of how to effectively handle multimodal optimization problems (i.e., the existence of multiple equivalent global optimal solutions) in multi-objective optimization. Existing multi-objective evolutionary algorithms (EMOA) face challenges in dealing with such problems because they need to maintain good convergence and diversity in both the objective space and the decision space simultaneously. Specifically, this paper proposes a new algorithm—Preference-Inspired Coevolutionary Algorithm with Active Diversity Strategy (MMPICEAg), aimed at solving multi-objective multimodal optimization problems.

### Main Contributions of the Paper:
1. **Proposed New Algorithm**: The MMPICEAg algorithm is proposed, which is based on the popular coevolutionary framework (PICEAg) and introduces diversity fitness assignment and dual diversity archive update strategies.
2. **Diversity Strategy**: By promoting diversity in both the objective space and the decision space simultaneously, the performance of the algorithm is improved.
3. **Experimental Validation**: The effectiveness of MMPICEAg is validated through comparative experiments with three general EMOAs and four state-of-the-art multimodal EMOAs.

### Background and Motivation:
- **Multi-Objective Optimization Problems (MOPs)**: Common in many practical applications, such as architectural design, aerospace mission design, rocket engine design, etc.
- **Multimodal Characteristics**: Many MOPs have multimodal characteristics, meaning that there are multiple Pareto optimal solutions in the decision space corresponding to the same point in the objective space.
- **Limitations of Existing Algorithms**: Existing EMOAs struggle to maintain good diversity and convergence in both the objective space and the decision space when dealing with multimodal MOPs.

### Solution:
- **MMPICEAg Algorithm**: Adopts a preference-inspired coevolutionary framework and introduces diversity fitness assignment and dual diversity archive update strategies.
- **Fitness Assignment**: Assigns fitness based on density ranking in the decision space, encouraging sparse candidate solutions to obtain higher fitness.
- **Archive Update**: Iteratively removes the most crowded solutions by considering the diversity of both the objective space and the decision space.

### Experimental Results:
- **Benchmark Testing**: Experiments conducted on three widely used benchmark test sets show that MMPICEAg outperforms competing algorithms on most problems.
- **Performance Advantage**: Particularly excels in decision space diversity.

### Conclusion:
The MMPICEAg algorithm performs excellently in handling multi-objective multimodal optimization problems and can serve as a benchmark algorithm in this field. Future research directions include further optimizing algorithm performance and exploring more practical applications.