Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
41 tokens/sec
o3 Pro
7 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Push and Pull Search for Solving Constrained Multi-objective Optimization Problems (1709.05915v1)

Published 15 Sep 2017 in cs.NE and cs.AI

Abstract: This paper proposes a push and pull search (PPS) framework for solving constrained multi-objective optimization problems (CMOPs). To be more specific, the proposed PPS divides the search process into two different stages, including the push and pull search stages. In the push stage, a multi-objective evolutionary algorithm (MOEA) is adopted to explore the search space without considering any constraints, which can help to get across infeasible regions very fast and approach the unconstrained Pareto front. Furthermore, the landscape of CMOPs with constraints can be probed and estimated in the push stage, which can be utilized to conduct the parameters setting for constraint-handling approaches applied in the pull stage. Then, a constrained multi-objective evolutionary algorithm (CMOEA) equipped with an improved epsilon constraint-handling is applied to pull the infeasible individuals achieved in the push stage to the feasible and non-dominated regions. Compared with other CMOEAs, the proposed PPS method can more efficiently get across infeasible regions and converge to the feasible and non-dominated regions by applying push and pull search strategies at different stages. To evaluate the performance regarding convergence and diversity, a set of benchmark CMOPs is used to test the proposed PPS and compare with other five CMOEAs, including MOEA/D-CDP, MOEA/D-SR, C-MOEA/D, MOEA/D-Epsilon and MOEA/D-IEpsilon. The comprehensive experimental results demonstrate that the proposed PPS achieves significantly better or competitive performance than the other five CMOEAs on most of the benchmark set.

Analyzing the Push and Pull Search Framework for Solving Constrained Multi-objective Optimization Problems

The paper presents the Push and Pull Search (PPS) framework, an innovative approach for addressing Constrained Multi-objective Optimization Problems (CMOPs). CMOPs are characterized by multiple conflicting objectives that must be optimized simultaneously under a set of constraint conditions. The novel aspect of this research lies in the division of the search process into two distinct stages: the push and the pull search stages.

PPS Framework: An Overview

The PPS framework introduces a bifurcated strategy for tackling CMOPs. In the initial push stage, constraints are disregarded as a Multi-objective Evolutionary Algorithm (MOEA) endeavors to navigate the search space swiftly. This phase aims to circumnavigate infeasible regions and draw near the unconstrained Pareto front. By probing the landscape during this stage, valuable information can be extracted to assist in parameter setting for the subsequent stage.

In the pull stage, on the other hand, the search is refined using a Constrained Multi-objective Evolutionary Algorithm (CMOEA) with improved epsilon constraint-handling. This stage concentrates on drawing infeasible solutions, identified in the push stage, into feasible and non-dominated regions.

Empirical Evaluation

The performance of the PPS method was rigorously tested against benchmark CMOPs, comparing it with five alternate CMOEAs: MOEA/D-CDP, MOEA/D-SR, C-MOEA/D, MOEA/D-Epsilon, and MOEA/D-IEpsilon. The PPS framework demonstrated superior or comparable performance on most benchmark CMOPs regarding both convergence and diversity. This was evidenced by statistical comparisons employing the well-established performance metrics: Inverted Generational Distance (IGD) and Hypervolume (HV).

Key Findings

From the empirical results, PPS showed significant efficiency in navigating large infeasible regions to approach unconstrained Pareto fronts during the push stage. The improved epsilon method effectively guided solutions into feasible regions during the pull stage. However, the parameter tuning between the stages remained critical, as indifferent tuning could impede performance, particularly on problems like LIR-CMOP9.

Implications and Future Work

The dual-stage strategy of PPS introduces a structured approach for handling constraints in CMOPs, with potential applicability to various complex, real-world optimization problems. The division of search into exploration (push) and exploitation (pull) phases can substantially enhance solution quality and computational efficiency, especially in high-dimensional problems with numerous constraints.

Future research directions may focus on integrating advanced constraint-handling mechanisms in the pull stage and employing machine learning techniques for improved parameter adaptation. Moreover, adapting the PPS paradigm to other frameworks beyond MOEA/D, such as NSGA-II, and testing against a broader spectrum of real-world problems could further validate its robustness and applicability.

The utility of parsing problem landscapes to inform search strategies, as demonstrated by PPS, underscores an important trend toward more informed and adaptive algorithms in evolutionary computation. This approach, when paired with sophisticated analytical techniques, can push the boundaries of what can be achieved in multi-objective optimization.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (8)
  1. Zhun Fan (29 papers)
  2. Wenji Li (17 papers)
  3. Xinye Cai (7 papers)
  4. Hui Li (1004 papers)
  5. Caimin Wei (5 papers)
  6. Qingfu Zhang (78 papers)
  7. Kalyanmoy Deb (42 papers)
  8. Erik D. Goodman (3 papers)
Citations (315)