pith. sign in

arxiv: 2604.03708 · v1 · submitted 2026-04-04 · 💻 cs.NE · cs.AI

RDEx-CMOP: Feasibility-Aware Indicator-Guided Differential Evolution for Fixed-Budget Constrained Multiobjective Optimization

Sichen Tao , Yifei Yang , Ruihan Zhao , Kaiyu Wang , Sicheng Liu , Shangce Gao This is my paper

Pith reviewed 2026-05-13 17:09 UTC · model grok-4.3

classification 💻 cs.NE cs.AI
keywords constrained multiobjective optimizationdifferential evolutionfeasibility-awareindicator-guidedCEC 2025 benchmarkfixed-budget optimizationevolutionary algorithms
0
0 comments X

The pith

RDEx-CMOP outperforms other algorithms on the CEC 2025 constrained multiobjective benchmark by achieving the highest total score and best average rank.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces RDEx-CMOP, a differential evolution algorithm designed for constrained multiobjective optimization problems where the number of evaluations is strictly limited. It combines three key elements: an epsilon-level schedule to guide the search toward feasible regions quickly, a SPEA2-style indicator to assign fitness based on convergence and diversity, and a mutation strategy that uses fitness to guide the current-to-pbest update. When tested on the official CEC 2025 CMOP benchmark with the median-target U-score, this combination yields the top overall score and rank among released competitors while maintaining near-zero constraint violations on most test problems. A sympathetic reader would care because many engineering and design tasks require balancing multiple objectives under constraints with limited computational resources, and improved algorithms can deliver better solutions in practice.

Core claim

RDEx-CMOP integrates an ε-level feasibility schedule, a SPEA2-style indicator-driven fitness assignment, and a fitness-oriented current-to-pbest/1 mutation operator to achieve superior performance in fixed-budget constrained multiobjective optimization, as demonstrated by its highest total score and best average rank on the CEC 2025 benchmark with strong target attainment and minimal final violations.

What carries the argument

The ε-level feasibility schedule together with SPEA2-style indicator-driven fitness assignment and fitness-oriented current-to-pbest/1 mutation operator, which together promote fast feasibility, stable convergence, and diversity preservation.

If this is right

  • RDEx-CMOP attains feasibility targets more effectively than competitors under the same evaluation budget.
  • It maintains near-zero final constraint violation on most benchmark problems.
  • The method secures the highest total score and best average rank among compared algorithms.
  • It balances convergence and diversity through indicator-based selection in constrained settings.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The approach could be adapted to other population-based methods for handling constraints in optimization.
  • Performance on the CEC benchmark suggests potential benefits in real-world applications like engineering design with multiple conflicting goals.
  • Future work might explore tuning the epsilon schedule for different problem types to further improve results.

Load-bearing premise

The median-target U-score and released trace data from the CEC 2025 CMOP benchmark accurately capture practical performance across a wide range of constrained multiobjective problems.

What would settle it

Running RDEx-CMOP and the comparison algorithms on a new collection of constrained multiobjective test problems with different characteristics and finding that it no longer ranks first or shows higher violations would falsify the performance claim.

read the original abstract

Constrained multiobjective optimisation requires fast feasibility attainment together with stable convergence and diversity preservation under strict evaluation budgets. This report documents RDEx-CMOP, the differential evolution variant used in the IEEE CEC 2025 numerical optimisation competition (C06 special session) constrained multiobjective track. RDEx-CMOP integrates an {\epsilon}-level feasibility schedule, a SPEA2-style indicator-driven fitness assignment, and a fitness-oriented current-to-pbest/1 mutation operator. We evaluate RDEx-CMOP on the official CEC 2025 CMOP benchmark using the median-target U-score framework and the released trace data. Experimental results show that RDEx-CMOP achieves the highest total score and the best overall average rank among all released comparison algorithms, with strong target-attainment behaviour and near-zero final violation on most problems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper introduces RDEx-CMOP, a differential evolution variant for fixed-budget constrained multiobjective optimization. It combines an ε-level feasibility schedule, SPEA2-style indicator-driven fitness assignment, and a current-to-pbest/1 mutation operator. Evaluated on the CEC 2025 CMOP benchmark via the median-target U-score framework and released trace data, the algorithm reports the highest total score and best average rank among released competitors, along with strong target attainment and near-zero final violations on most problems.

Significance. If the reported ranking holds under scrutiny, the work supplies a competitive baseline for the IEEE CEC 2025 constrained multiobjective track. The explicit integration of established mechanisms (ε-level scheduling and SPEA2 indicator) into a DE framework with fitness-oriented mutation offers a practical, internally consistent approach for feasibility-aware optimization under strict evaluation limits.

major comments (1)
  1. §4 (Experimental Results): the claim of highest total score and best average rank is presented without statistical significance tests (e.g., Friedman or Wilcoxon rank-sum) or release of raw performance vectors, rendering independent verification of the ranking impossible from the supplied material.
minor comments (2)
  1. Abstract and §3 (Algorithm Description): the ε-level schedule and indicator parameters are described at a high level; explicit pseudocode or default values would improve reproducibility.
  2. §4: the manuscript should state whether the released trace data will be archived with the paper or competition repository.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive assessment of RDEx-CMOP and the recommendation for minor revision. We address the single major comment on the experimental results below.

read point-by-point responses

  1. Referee: §4 (Experimental Results): the claim of highest total score and best average rank is presented without statistical significance tests (e.g., Friedman or Wilcoxon rank-sum) or release of raw performance vectors, rendering independent verification of the ranking impossible from the supplied material.

    Authors: We agree that the current presentation would benefit from explicit statistical validation. Although the CEC 2025 CMOP track uses the median-target U-score as the official ranking metric and trace data have already been released, we will strengthen §4 by adding a Friedman test across all benchmark problems followed by post-hoc Wilcoxon rank-sum tests with Holm correction. We will also deposit the complete raw performance vectors (per run, per problem) as supplementary material linked from the paper to enable full independent verification. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper presents an algorithmic description of RDEx-CMOP (ε-level schedule, SPEA2-style indicator, current-to-pbest/1 operator) followed by empirical ranking on the external CEC 2025 CMOP benchmark under the median-target U-score. No derivations, fitted parameters renamed as predictions, or load-bearing self-citations appear; the central claim rests on standard benchmark evaluation rather than internal reduction to inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no free parameters, axioms, or invented entities are extractable or required for the high-level claim.

pith-pipeline@v0.9.0 · 5456 in / 1005 out tokens · 30192 ms · 2026-05-13T17:09:46.451968+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages

  1. [1]

    A fast and elitist multiobjective genetic algorithm: NSGA-II,

    K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, “A fast and elitist multiobjective genetic algorithm: NSGA-II,”IEEE Transactions on Evolutionary Computation, vol. 6, no. 2, pp. 182–197, 2002

    work page 2002

  2. [2]

    SPEA2: Improving the strength Pareto evolutionary algorithm,

    E. Zitzler, M. Laumanns, and L. Thiele, “SPEA2: Improving the strength Pareto evolutionary algorithm,” inEvolutionary Methods for Design, Optimisation and Control with Applications to Industrial Problems (EUROGEN 2001), 2001, pp. 95–100

    work page 2001

  3. [3]

    Indicator-based selection in multiobjective search,

    E. Zitzler and S. K ¨unzli, “Indicator-based selection in multiobjective search,” inParallel Problem Solving from Nature – PPSN VIII. Springer, 2004, pp. 832–842

    work page 2004

  4. [4]

    MOEA/D: A multiobjective evolutionary algorithm based on decomposition,

    Q. Zhang and H. Li, “MOEA/D: A multiobjective evolutionary algorithm based on decomposition,”IEEE Transactions on Evolutionary Compu- tation, vol. 11, no. 6, pp. 712–731, 2007

    work page 2007

  5. [5]

    An efficient constraint handling method for genetic algorithms,

    K. Deb, “An efficient constraint handling method for genetic algorithms,” Computer Methods in Applied Mechanics and Engineering, vol. 186, no. 2-4, pp. 311–338, 2000

    work page 2000

  6. [6]

    Constrained optimization by theεcon- strained differential evolution with gradient-based mutation and feasible elites,

    T. Takahama and S. Sakai, “Constrained optimization by theεcon- strained differential evolution with gradient-based mutation and feasible elites,” in2006 IEEE International Conference on Evolutionary Com- putation. IEEE, 2006, pp. 1–8

    work page 2006

  7. [7]

    Constraint-handling in nature-inspired numerical optimization: Past, present and future,

    E. Mezura-Montes and C. A. Coello Coello, “Constraint-handling in nature-inspired numerical optimization: Past, present and future,”Swarm and Evolutionary Computation, vol. 1, no. 4, pp. 173–194, 2011

    work page 2011

  8. [8]

    An evolutionary multitasking optimization framework for constrained multiobjective op- timization problems,

    K. Qiao, K. Yu, B. Qu, J. Liang, H. Song, and C. Yue, “An evolutionary multitasking optimization framework for constrained multiobjective op- timization problems,”IEEE Transactions on Evolutionary Computation, vol. 26, no. 5, pp. 1062–1076, 2022

    work page 2022

  9. [9]

    Dynamic auxiliary task-based evolutionary multitasking for constrained multiobjective optimization,

    K. Qiao, K. Yu, B. Qu, J. Liang, H. Song, C. Yue, H. Lin, and K. C. Tan, “Dynamic auxiliary task-based evolutionary multitasking for constrained multiobjective optimization,”IEEE Transactions on Evolutionary Computation, vol. 27, no. 6, pp. 1507–1521, 2023

    work page 2023

  10. [10]

    Evolutionary constrained multiobjective optimization: Scalable high- dimensional constraint benchmarks and algorithm,

    K. Qiao, J. Liang, K. Yu, C. Yue, H. Lin, D. Zhang, and B. Qu, “Evolutionary constrained multiobjective optimization: Scalable high- dimensional constraint benchmarks and algorithm,”IEEE Transactions on Evolutionary Computation, vol. 27, no. 6, pp. 1667–1681, 2023

    work page 2023

  11. [11]

    Con- straint subsets-based evolutionary multitasking for constrained multiob- jective optimization,

    K. Yu, L. Wang, J. Liang, H. Wang, K. Qiao, and T. Liang, “Con- straint subsets-based evolutionary multitasking for constrained multiob- jective optimization,”Swarm and Evolutionary Computation, vol. 86, p. 101531, 2024

    work page 2024

  12. [12]

    A dynamic ex- emplars selection-based differential evolution algorithm for constrained multi-objective optimization,

    X. Ban, J. Liang, K. Yu, K. Qiao, Y . Wang, and J. Peng, “A dynamic ex- emplars selection-based differential evolution algorithm for constrained multi-objective optimization,” in2024 International Symposium on Algorithmic Software Engineering. IEEE, 2024, pp. 1–8

    work page 2024

  13. [13]

    Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces,

    R. Storn and K. Price, “Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces,”Journal of Global Optimization, vol. 11, pp. 341–359, 1997

    work page 1997

  14. [14]

    JADE: adaptive differential evolution with optional external archive,

    J. Zhang and A. C. Sanderson, “JADE: adaptive differential evolution with optional external archive,”IEEE Transactions on Evolutionary Computation, vol. 13, no. 5, pp. 945–958, 2009

    work page 2009

  15. [15]

    Success-history based parameter adapta- tion for differential evolution,

    R. Tanabe and A. Fukunaga, “Success-history based parameter adapta- tion for differential evolution,” in2013 IEEE Congress on Evolutionary Computation (CEC). IEEE, 2013, pp. 71–78

    work page 2013

  16. [16]

    Improving the search performance of SHADE using linear population size reduction,

    R. Tanabe and A. S. Fukunaga, “Improving the search performance of SHADE using linear population size reduction,” in2014 IEEE Congress on Evolutionary Computation (CEC). IEEE, 2014, pp. 1658–1665

    work page 2014

  17. [17]

    Trial-based dominance for comparing both the speed and accuracy of stochastic optimizers with standard non-parametric tests,

    K. V . Price, A. Kumar, and P. N. Suganthan, “Trial-based dominance for comparing both the speed and accuracy of stochastic optimizers with standard non-parametric tests,”Swarm and Evolutionary Computation, vol. 78, p. 101287, 2023

    work page 2023

  18. [18]

    2025 CEC Competition Repository,

    P. N. Suganthan, “2025 CEC Competition Repository,” GitHub repository, 2025, accessed: 2026-03-09. [Online]. Available: https: //github.com/P-N-Suganthan/2025-CEC APPENDIXA SUPPLEMENTARYU-SCORETABLES TABLE IV CEC 2025 CMOPEVALUATION(MEDIAN TARGET):AVERAGE RANKINGS OVER15PROBLEMS(LOWER IS BETTER)FOR ALL RELEASED COMPARISON ALGORITHMS. Rank Algorithm Total ...

    work page 2025