Pith. sign in

REVIEW 3 major objections 5 minor 54 references

An Efficient Evolutionary Algorithm for Few-for-Many Optimization

T0 review · 3 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read SoM-EMOA, an evolutionary algorithm that directly optimizes the sum-of-minimum coverage objective, outperforms existing few-for-many solvers on 31 of 32 synthetic instances and on both real-world problems.

desk verdict A genuinely useful F4M algorithm with a sound design, but the "superior for many objectives" claim rests on a benchmark built from the algorithm's own objective, so the empirical case is weaker than the abstract suggests. read the letter →

arxiv 2601.06387 v1 pith:E6WD76AM submitted 2026-01-10 cs.NE

Ke Shang , Hisao Ishibuchi , Zexuan Zhu , Qingfu Zhang This is my paper · ORCID
classification cs.NE
keywords Few-for-manyoptimizationMulti-objectiveMany-objectiveEvolutionaryalgorithmSum-of-minimumobjectiveR2indicatorBenchmarktestsuiteCoverage
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that few-for-many optimization—finding k solutions that jointly cover m objectives, where k is much smaller than m—is best attacked by directly optimizing the set-level coverage objective, and that a simple (μ+1) evolutionary strategy can do this efficiently. Its algorithm SoM-EMOA removes the solution whose exclusion damages coverage least, and uses an archive of per-objective bests to guide offspring toward currently weak objectives. To enable evaluation, the paper constructs a benchmark suite that turns any multi-objective problem into an F4M instance via Tchebycheff scalarization, exploiting the formal similarity between the F4M objective and the R2 indicator. Across 32 synthetic instances and two real-world problems, SoM-EMOA reports the lowest mean coverage value on 31 synthetic instances and on both real-world problems.

What carries the argument

The central object is the sum-of-minimum set objective G_ws(X_k) = Σ_{i=1}^m min_{1≤j≤k} f_i(x^(j)); it defines what "covering" means and is the selection criterion for removal. The archive A stores per-objective best solutions and drives mating through the probability vector p = (v−u)/||v−u||_1, where v is the population's current per-objective minima and u is the archive's, biasing offspring toward weak objectives. The benchmark is built with the Tchebycheff scalarization F_i(x) = max_j w^(i)_j |f_j(x) − z*_j|, which is the formal link between F4M and the R2 indicator.

What would settle it

Construct an independent set of F4M instances with m = 100 uncorrelated objectives (e.g., each objective's optimum at a separate decision-variable coordinate) and run SoM-EMOA against a front-approximation method followed by greedy subset selection under identical evaluation budgets; if SoM-EMOA does not achieve the lowest mean G_ws over 30 runs, the claimed advantage is benchmark-specific.

Watch

Extended reading notes

Core claim

Few-for-many (F4M) optimization asks for a small set X_k of k solutions that collectively cover m objectives, with coverage measured by G_ws(X_k) = Σ_i min_{j≤k} f_i(x^(j)). SoM-EMOA is a (μ+1) evolutionary strategy that uses exactly this set-level objective for selection: after adding an offspring, it removes the solution whose exclusion least degrades G_ws; offspring generation is biased by an archive that stores the best value seen for each objective, so poorly covered objectives get targeted. The paper also introduces a benchmark suite that transforms any multi-objective problem into an F4M instance by scalarizing m Tchebycheff objectives, and reports that SoM-EMOA attains the best mean

Load-bearing premise

The load-bearing premise is that the new R2-based benchmark suite—built by scalarizing low-dimensional problems with the same sum-of-minimum objective SoM-EMOA optimizes—captures the difficulty of real few-for-many problems; if it does not, the reported superiority may be a self-consistent demonstration rather than evidence about the real world.

Editorial extensions

If this is right

  • If the central claim holds, small chosen sets of solutions—far smaller than the number of objectives—can replace large Pareto-front approximations when many objectives are correlated.
  • The benchmark construction means any existing multi-objective test problem can be turned into a controlled F4M instance, supporting systematic studies of how algorithm performance scales with m and k.
  • Because SoM-EMOA relies only on function evaluations, it extends F4M optimization to black-box and combinatorial settings where gradient-based sum-of-minimum methods do not apply.
  • The reported results imply that optimizing the set-level coverage objective directly is more effective than first approximating the entire Pareto front and then selecting k solutions from it, at least in the tested settings.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: The new benchmark is generated from the same objective SoM-EMOA directly optimizes, so independent validation with uncorrelated objectives would test whether the 31-of-32 edge reflects a genuine advantage or alignment between the algorithm and the benchmark.
  • Editorial inference: The per-objective archive plus coverage-deficiency mating can be read as a general set-search mechanism, potentially transferable to set-based Bayesian optimization or multi-task settings where a few models must cover many tasks.
  • Editorial inference: The close formal link between the F4M objective and the R2 indicator suggests that advances in R2-optimal subset selection could directly improve F4M solvers, and vice versa; this cross-fertilization is not exploited in the paper itself.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes SoM-EMOA, a (μ+1) evolutionary strategy for few-for-many (F4M) optimization. The algorithm maintains a small population of k solutions and an archive of per-objective best solutions; offspring are generated by mating a population member with an archive member chosen according to a probability vector proportional to the gap between population and archive minima; after adding the offspring, the solution whose removal most improves the sum-of-minimum coverage G_ws is removed, with an O(mk) average-case implementation of this removal rule. The paper also introduces an R2-based benchmark construction in which any q-objective MOP is scalarized into m Tchebycheff objectives, and G_ws is used as the F4M objective. Experiments compare SoM-EMOA with CluSO, MOCOBO, and several EMO baselines on DC-MaTS/NMLR, on the new F4M-DTLZ/WFG suite with m = 25,...,100 and k = 5, and on DDMOP1/4. The reported results show SoM-EMOA best on all 5 existing synthetic problems, 31 of 32 new benchmark instances, and best or statistically comparable on the real-world problems.

Significance. F4M/MaCP is a timely and understudied problem, and the algorithmic core of this work is conceptually clean. The removal rule in Section III-E exactly selects the best size-k subset of the temporary k+1 population, the archive mechanism preserves per-objective elitism, and the mating-selection idea is intuitive. Source code is promised. Several results, especially on DC-MaTS and NMLR (Table I), show large and consistent margins over existing methods, which is meaningful independent evidence for the algorithm's value. However, the headline claim of superiority 'especially on instances involving a large number of objectives' rests primarily on a new benchmark that is constructed from the very sum-of-minimum objective that SoM-EMOA directly optimizes, and the 'many-objective' instances are derived from only 3 base objectives. The empirical comparison is therefore not yet an independent validation of the central claim, even though the algorithm is plausible and the non-benchmark results are encouraging.

major comments (3)
  1. [Section IV-C, Eqs. (12)-(13); cf. Eq. (3) and Algorithms 2-4] The proposed benchmark defines m scalarized objectives via Tchebycheff weights and then evaluates algorithms by exactly the sum-of-minimum G_ws that SoM-EMOA is engineered to minimize. This is a self-consistent loop: the algorithm directly optimizes G_ws through its removal step (Algorithm 3) and archive-guided mating, while the EMO baselines are not designed for G_ws and are post-processed by greedy subset selection (Algorithm 4). Consequently, Table II largely demonstrates that SoM-EMOA can optimize its own objective on these constructed instances, not that it finds better F4M solutions in an independent sense. To support the superiority claim, the authors should either validate the new benchmark against known optimal cover sets for at least some instances, include F4M problems whose objectives are not generated by the same scalarization SoM-EMOA minimizes, or base the high-dimensional
  2. [Section V-A, Table II; Section V-B] All F4M-DTLZ and F4M-WFG instances are generated from q = 3 base problems, so the underlying Pareto front is at most two-dimensional; m = 25,...,100 merely re-scalarizes a few trade-off directions. With k = 5, a generic front-approximation algorithm can already give high coverage after greedy selection, as the paper itself concedes in Section V-B. The observed margins are tiny (often well below 1%) and not monotonic in m; in F4M-DTLZ4 at m = 100, NSGA-II, PREA and NRV-MOEA are statistically better than SoM-EMOA. Thus the abstract's claim of superiority 'especially on instances involving a large number of objectives' is not supported by the new benchmark. The claim should be removed or qualified, or tested with base problems of substantially higher q (e.g., q >= 10) so that the intrinsic trade-off dimension is comparable to the claimed high-dimensional setting.
  3. [Section V-C, DDMOP1/DDMOP4] The real-world validation uses problems with only 9 and 10 objectives, which does not match the 'large number of objectives' emphasis of the paper. On DDMOP4 at 1000 FEs, CluSO has the best mean G_ws and SoM-EMOA is only statistically comparable. These experiments provide useful evidence in modest-dimensional settings, but they do not compensate for the weakness in the high-dimensional synthetic evidence. The authors should either add real-world or data-driven F4M problems with many genuinely independent objectives, or explicitly restrict the superiority claim to the tested range.
minor comments (5)
  1. [Section III-C, Eq. (8)] The probability vector p = (v-u)/||v-u||_1 is undefined when v = u (i.e., when the archive and current population have identical per-objective minima). A safeguard such as uniform sampling should be specified.
  2. [Tables I-III] The experimental section states that mean and standard deviation are reported, but the tables list only mean values. Either include the standard deviations or state clearly that they are omitted for brevity.
  3. [Table II, F4M-DTLZ4 m=100] The aggregate claim of 'best on 31 of 32 instances' is accompanied by three '+' entries in the same row (NSGA-II, PREA, NRV-MOEA). The narrative should explicitly acknowledge these exceptions and discuss what they imply about the benchmark's discriminative power.
  4. [Section V-D] MOCOBO is excluded from the synthetic and 10,000-FE comparisons because of runtime, but this is only stated in Section V-D after Table III. The limitation should be disclosed in the experimental setup.
  5. [Algorithm 3] In the pseudo-code, v' is updated inside the loop over x for i not in I, but it is not explicitly reset for each candidate x. Clarify that v' is a temporary vector computed per candidate.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the R2-based benchmark is aligned with SoM-EMOA's objective by design, but the central claim is independently supported by pre-existing benchmarks.

full rationale

The paper's derivation chain is not circular. SoM-EMOA directly optimizes the sum-of-minimum objective G_ws in Eq. (3), and the proposed R2-based benchmark in Section IV constructs F4M instances whose objective (Eq. 13) is the same sum-of-minimum form applied to scalarized objectives (Eq. 12). This alignment is explicit and disclosed: the benchmark is "built upon the connection between the R2 indicator and the objective of F4M optimization," so it is by construction a direct instantiation of the F4M formulation rather than a hidden fitted input. The algorithm's design choices are evaluated by ablations, not by fitting parameters to test outcomes, and the paper does not rename a known result: it cites prior work on sum-of-minimum and R2 indicators transparently. Crucially, the empirical claim of superior performance is not based solely on the authors' new benchmark. Table I reports results on DC-MaTS and NMLR, which are pre-existing benchmarks introduced in prior work [13], [14], and Table III reports real-world DDMOP1/DDMOP4 problems from an external repository [45]. These benchmarks were not constructed for SoM-EMOA and provide independent grounding. The paper itself notes in Section V-B that the low-dimensional base problems (q=3) can make coverage easy for standard EMO algorithms, which is a benchmark-validity limitation, not a circular derivation. Self-citations to [13] and [14] involve a shared author, but the cited benchmarks and problem definitions are pre-existing, externally available, and not tailored to this paper's algorithm. No uniqueness theorem or ansatz is imported from the authors' prior work to force a choice. Therefore, there is no circular step by construction or by fitted-input renaming.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical or mathematical entities are introduced. The algorithm's archive and probability distribution are algorithmic constructs, not independent postulated entities. The free parameters are standard evolutionary algorithm hyperparameters plus the unspecified initial archive size N. The benchmark generator relies on standard scalarization properties and the R2 indicator. The central claim depends on the domain assumption that G_ws is the right F4M objective and that the R2-based benchmark represents realistic F4M problems.

free parameters (5)
  • N (initial archive set size) = not specified in paper
    Algorithm 1 initializes a large set S of size N≫k, but the experimental section does not state N. This parameter affects archive initialization quality and total function evaluation budget.
  • SBX distribution index η_c = 20
    Chosen from prior work [13]; not tuned in this paper, but affects offspring generation.
  • Polynomial mutation index η_m = 20
    Chosen from prior work [13]; affects mutation step sizes.
  • Crossover probability p_c = 1.0
    Set following [13]; standard for SBX.
  • Mutation probability p_m = 1/d (d = number of decision variables)
    Standard polynomial mutation probability; follows [13].
assumptions (4)
  • standard math The Tchebycheff scalarization of a multi-objective problem has its optimal solutions on the Pareto front of the original problem.
    Used in Section IV.C to construct the benchmark: each F4M objective F_i(x) = g_tch(x|w^(i), z*) is claimed to have optima on the base Pareto front. This is a well-known property of Tchebycheff scalarization for convex/continuous problems.
  • domain assumption The sum-of-minimum objective G_ws(X_k) is the appropriate formulation of few-for-many optimization.
    Section II.B adopts G_ws as 'the' F4M objective, citing Liu et al. [13] and Lin et al. [14]. The entire algorithm and benchmark are built around this choice; other formulations (e.g., Tchebycheff set scalarization) are explicitly deferred to future work.
  • domain assumption The R2 indicator is a valid measure of set-level coverage quality.
    Section IV.B motivates the benchmark via the R2 indicator; the paper assumes that optimizing G_ws (≈ m×R2) captures what it means for a small set to 'effectively handle' many objectives.
  • domain assumption The archive of per-objective best solutions (Algorithm 1, Eq. 5) provides useful guidance for offspring generation.
    Section III.B/C asserts that selecting archive solutions with probability proportional to v-u improves search. No proof is given; ablation studies in Section V.E provide empirical support, but the assumption is not theoretically grounded.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Efficient Evolutionary Algorithm for Few-for-Many Optimization." pith.science (2026). https://pith.science/paper/E6WD76AM

@misc{pith2026260106387,
  author       = {Pith},
  title        = {Pith review of: An Efficient Evolutionary Algorithm for Few-for-Many Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/E6WD76AM}},
  note         = {Machine review of arXiv:2601.06387}
}
abstract

Few-for-many (F4M) optimization, recently introduced as a novel paradigm in multi-objective optimization, aims to find a small set of solutions that effectively handle a large number of conflicting objectives. Unlike traditional many-objective optimization methods, which typically attempt comprehensive coverage of the Pareto front, F4M optimization emphasizes finding a small representative solution set to efficiently address high-dimensional objective spaces. Motivated by the computational complexity and practical relevance of F4M optimization, this paper proposes a new evolutionary algorithm explicitly tailored for efficiently solving F4M optimization problems. Inspired by SMS-EMOA, our proposed approach employs a $(\mu+1)$-evolution strategy guided by the objective of F4M optimization. Furthermore, to facilitate rigorous performance assessment, we propose a novel benchmark test suite specifically designed for F4M optimization by leveraging the similarity between the R2 indicator and F4M formulations. Our test suite is highly flexible, allowing any existing multi-objective optimization problem to be transformed into a corresponding F4M instance via scalarization using the weighted Tchebycheff function. Comprehensive experimental evaluations on benchmarks demonstrate the superior performance of our algorithm compared to existing state-of-the-art algorithms, especially on instances involving a large number of objectives. The source code of the proposed algorithm will be released publicly. Source code is available at https://github.com/MOL-SZU/SoM-EMOA.

Figures

Figures reproduced from arXiv: 2601.06387 by the authors.

Figure 1
Figure 1. An illustration of F4M optimization with [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the F4M formulation on the 2-objective DTLZ2 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. The convergence curve of different algorithms on 5 existing F4M [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The convergence curve of different algorithms on the proposed F4M optimization benchmark problems. The number of objectives [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Comparison of runtime (in seconds) for different algorithms [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 3 linked inside Pith

  1. [1]

    A fast and eli- tist multiobjective genetic algorithm: NSGA-II,

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

    2002

  2. [2]

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

    Q. Zhang and H. Li, “MOEA/D: A multiobjective evolutionary algorithm based on decomposition,”IEEE T ransactions on Evolu- tionary Computation, vol. 11, no. 6, pp. 712–731, 2007

    2007

  3. [3]

    An evolutionary many-objective optimiza- tion algorithm using reference-point-based nondominated sorting approach, part I: solving problems with box constraints,

    K. Deb and H. Jain, “An evolutionary many-objective optimiza- tion algorithm using reference-point-based nondominated sorting approach, part I: solving problems with box constraints,”IEEE T ransactions on Evolutionary Computation, vol. 18, no. 4, pp. 577– 601, 2013

    2013

  4. [4]

    A reference vector guided evolutionary algorithm for many-objective optimization,

    R. Cheng, Y. Jin, M. Olhofer, and B. Sendhoff, “A reference vector guided evolutionary algorithm for many-objective optimization,” IEEE T ransactions on Evolutionary Computation, vol. 20, no. 5, pp. 773–791, 2016

    2016

  5. [5]

    Deep reinforcement learning based adaptive operator selection for evolutionary multi-objective optimization,

    Y. Tian, X. Li, H. Ma, X. Zhang, K. C. Tan, and Y. Jin, “Deep reinforcement learning based adaptive operator selection for evolutionary multi-objective optimization,”IEEE T ransactions on Emerging T opics in Computational Intelligence, vol. 7, no. 4, pp. 1051– 1064, 2022

    2022

  6. [6]

    Direct preference-based evolution- ary multi-objective optimization with dueling bandits,

    T. Huang, S. Wang, and K. Li, “Direct preference-based evolution- ary multi-objective optimization with dueling bandits,”Advances in Neural Information Processing Systems, vol. 37, pp. 122 206– 122 258, 2024

    2024

  7. [7]

    A survey of evolutionary algorithms for multi-objective optimization problems with irreg- ular Pareto fronts,

    Y. Hua, Q. Liu, K. Hao, and Y. Jin, “A survey of evolutionary algorithms for multi-objective optimization problems with irreg- ular Pareto fronts,”IEEE/CAA Journal of Automatica Sinica, vol. 8, no. 2, pp. 303–318, 2021

    2021

  8. [8]

    A grid-based evolutionary algorithm for many-objective optimization,

    S. Yang, M. Li, X. Liu, and J. Zheng, “A grid-based evolutionary algorithm for many-objective optimization,”IEEE T ransactions on Evolutionary Computation, vol. 17, no. 5, pp. 721–736, 2013

    2013

Show all 54 references
  1. [9]

    Diversity assessment in many- objective optimization,

    H. Wang, Y. Jin, and X. Yao, “Diversity assessment in many- objective optimization,”IEEE T ransactions on Cybernetics, vol. 47, no. 6, pp. 1510–1522, 2016

    2016

  2. [10]

    Evolutionary many- objective optimization: A short review,

    H. Ishibuchi, N. Tsukamoto, and Y. Nojima, “Evolutionary many- objective optimization: A short review,” in2008 IEEE Congress on Evolutionary Computation. IEEE, 2008, pp. 2419–2426

    2008

  3. [11]

    Many-objective prob- lems: challenges and methods,

    A. L ´opez Jaimes and C. A. Coello Coello, “Many-objective prob- lems: challenges and methods,”Springer Handbook of Computa- tional Intelligence, pp. 1033–1046, 2015

    2015

  4. [12]

    Evolutionary many-objective optimiza- tion: Difficulties, approaches, and discussions,

    H. Sato and H. Ishibuchi, “Evolutionary many-objective optimiza- tion: Difficulties, approaches, and discussions,”IEEJ T ransactions on Electrical and Electronic Engineering, vol. 18, no. 7, pp. 1048– 1058, 2023. 15

    2023

  5. [13]

    Many-objective cover problem: Discovering few solutions to cover many objectives,

    Y. Liu, C. Lu, X. Lin, and Q. Zhang, “Many-objective cover problem: Discovering few solutions to cover many objectives,” inInternational Conference on Parallel Problem Solving from Nature. Springer, 2024, pp. 68–82

    2024

  6. [14]

    Few for many: Tchebycheff set scalarization for many-objective optimiza- tion,

    X. Lin, Y. Liu, X. Zhang, F. Liu, Z. Wang, and Q. Zhang, “Few for many: Tchebycheff set scalarization for many-objective optimiza- tion,”arXiv preprint arXiv:2405.19650, 2024

    2024

  7. [15]

    Behavior of multi- objective evolutionary algorithms on many-objective knapsack problems,

    H. Ishibuchi, N. Akedo, and Y. Nojima, “Behavior of multi- objective evolutionary algorithms on many-objective knapsack problems,”IEEE T ransactions on Evolutionary Computation, vol. 19, no. 2, pp. 264–283, 2014

    2014

  8. [16]

    Objective reduc- tion in many-objective optimization: evolutionary multiobjective approaches and comprehensive analysis,

    Y. Yuan, Y.-S. Ong, A. Gupta, and H. Xu, “Objective reduc- tion in many-objective optimization: evolutionary multiobjective approaches and comprehensive analysis,”IEEE T ransactions on Evolutionary Computation, vol. 22, no. 2, pp. 189–210, 2017

    2017

  9. [17]

    Many-objective multi-solution transport,

    Z. Li, T. Li, V . Smith, J. Bilmes, and T. Zhou, “Many-objective multi-solution transport,”arXiv preprint arXiv:2403.04099, 2024

    2024 arXiv

  10. [18]

    Multi- and many-objective optimization: present and future in de novo drug design,

    J. S. Angelo, I. A. Guedes, H. J. Barbosa, and L. E. Dardenne, “Multi- and many-objective optimization: present and future in de novo drug design,”Frontiers in Chemistry, vol. 11, p. 1288626, 2023

    2023

  11. [19]

    Psychologi- cal targeting as an effective approach to digital mass persuasion,

    S. C. Matz, M. Kosinski, G. Nave, and D. J. Stillwell, “Psychologi- cal targeting as an effective approach to digital mass persuasion,” Proceedings of the National Academy of Sciences, vol. 114, no. 48, pp. 12 714–12 719, 2017

    2017

  12. [20]

    Estimating peer effects in networks with peer encouragement designs,

    D. Eckles, R. F. Kizilcec, and E. Bakshy, “Estimating peer effects in networks with peer encouragement designs,”Proceedings of the National Academy of Sciences, vol. 113, no. 27, pp. 7316–7322, 2016

    2016

  13. [21]

    Which tasks should be learned together in multi-task learning?

    T. Standley, A. Zamir, D. Chen, L. Guibas, J. Malik, and S. Savarese, “Which tasks should be learned together in multi-task learning?” inInternational Conference on Machine Learning. PMLR, 2020, pp. 9120–9132

    2020

  14. [22]

    Mot- ley: Benchmarking heterogeneity and personalization in federated learning,

    S. Wu, T. Li, Z. Charles, Y. Xiao, Z. Liu, Z. Xu, and V . Smith, “Mot- ley: Benchmarking heterogeneity and personalization in federated learning,”arXiv preprint arXiv:2206.09262, 2022

    2022 arXiv

  15. [23]

    Multi-objective coverage bayesian optimization (mocobo),

    N. Maus, K. Kim, Y. Zeng, H. T. Jones, F. Wan, M. D. T. Torres, C. de la Fuente-Nunez, and J. R. Gardner, “Multi-objective coverage bayesian optimization (mocobo),” 2025. [Online]. Available: https://arxiv.org/abs/2501.19342

    2025

  16. [24]

    Miettinen,Nonlinear multiobjective optimization

    K. Miettinen,Nonlinear multiobjective optimization. Springer Sci- ence & Business Media, 1999, vol. 12

    1999

  17. [25]

    Multi-objective optimiza- tion,

    K. Deb, K. Sindhya, and J. Hakanen, “Multi-objective optimiza- tion,” inDecision Sciences. CRC Press, 2016, pp. 161–200

    2016

  18. [26]

    Bridging the gap: Many-objective optimization and informed decision- making,

    K. S. Bhattacharjee, H. K. Singh, M. Ryan, and T. Ray, “Bridging the gap: Many-objective optimization and informed decision- making,”IEEE T ransactions on Evolutionary Computation, vol. 21, no. 5, pp. 813–820, 2017

    2017

  19. [27]

    Improving vi- sualization design for effective multi-objective decision making,

    B. Dy, N. Ibrahim, A. Poorthuis, and S. Joyce, “Improving vi- sualization design for effective multi-objective decision making,” IEEE T ransactions on Visualization and Computer Graphics, vol. 28, no. 10, pp. 3405–3416, 2021

    2021

  20. [28]

    An exact algorithm for a class of geometric set-cover problems,

    C. Contardo and A. Hertz, “An exact algorithm for a class of geometric set-cover problems,”Discrete Applied Mathematics, vol. 300, pp. 25–35, 2021

    2021

  21. [29]

    Applying the quantum approximate optimization algorithm to the minimum vertex cover problem,

    Y. Zhang, X. Mu, X.-W. Liu, X. Wang, X. Zhang, K. Li, T. Wu, D. Zhao, and C. Dong, “Applying the quantum approximate optimization algorithm to the minimum vertex cover problem,” Applied Soft Computing, vol. 118, p. 108554, 2022

    2022

  22. [30]

    Efficient algorithms for sum-of-minimum optimization,

    L. Ding, Z. Chen, X. Wang, and W. Yin, “Efficient algorithms for sum-of-minimum optimization,” inProceedings of the 41st International Conference on Machine Learning, 2024, pp. 10 927– 10 959

    2024

  23. [31]

    Pareto multi-task learning,

    X. Lin, H.-L. Zhen, Z. Li, Q.-F. Zhang, and S. Kwong, “Pareto multi-task learning,”Advances in Neural Information Processing Systems, vol. 32, 2019

    2019

  24. [32]

    Learning how to ask: Querying LMs with mixtures of soft prompts,

    G. Qin and J. Eisner, “Learning how to ask: Querying LMs with mixtures of soft prompts,”arXiv preprint arXiv:2104.06599, 2021

    2021 arXiv

  25. [33]

    Black-box optimization: Methods and applications,

    I. Bajaj, A. Arora, and M. F. Hasan, “Black-box optimization: Methods and applications,” inBlack Box Optimization, Machine Learning, and No-free Lunch Theorems. Springer, 2021, pp. 35–65

    2021

  26. [34]

    B. H. Korte, J. Vygen, B. Korte, and J. Vygen,Combinatorial optimization. Springer, 2011, vol. 1

    2011

  27. [35]

    SMS-EMOA: Multi- objective selection based on dominated hypervolume,

    N. Beume, B. Naujoks, and M. Emmerich, “SMS-EMOA: Multi- objective selection based on dominated hypervolume,”European Journal of Operational Research, vol. 181, no. 3, pp. 1653–1669, 2007

    2007

  28. [36]

    Parego: A hybrid algorithm with on-line landscape approximation for expensive multiobjective optimization prob- lems,

    J. Knowles, “Parego: A hybrid algorithm with on-line landscape approximation for expensive multiobjective optimization prob- lems,”IEEE transactions on evolutionary computation, vol. 10, no. 1, pp. 50–66, 2006

    2006

  29. [37]

    Scalable test problems for evolutionary multiobjective optimization,

    K. Deb, L. Thiele, M. Laumanns, and E. Zitzler, “Scalable test problems for evolutionary multiobjective optimization,” inEvolu- tionary Multiobjective Optimization: Theoretical Advances and Appli- cations. Springer, 2005, pp. 105–145

    2005

  30. [38]

    A review of multiobjective test problems and a scalable test problem toolkit,

    S. Huband, P . Hingston, L. Barone, and L. While, “A review of multiobjective test problems and a scalable test problem toolkit,” IEEE T ransactions on Evolutionary Computation, vol. 10, no. 5, pp. 477–506, 2006

    2006

  31. [39]

    Per- formance of decomposition-based many-objective algorithms strongly depends on pareto front shapes,

    H. Ishibuchi, Y. Setoguchi, H. Masuda, and Y. Nojima, “Per- formance of decomposition-based many-objective algorithms strongly depends on pareto front shapes,”IEEE T ransactions on Evolutionary Computation, vol. 21, no. 2, pp. 169–190, 2016

    2016

  32. [40]

    Exact calculation and proper- ties of the R2 multiobjective quality indicator,

    A. Jaszkiewicz and P . Zielniewicz, “Exact calculation and proper- ties of the R2 multiobjective quality indicator,”IEEE T ransactions on Evolutionary Computation, 2024

    2024

  33. [41]

    Reinvestigating the R2 indica- tor: Achieving Pareto compliance by integration,

    L. Sch ¨apermeier and P . Kerschke, “Reinvestigating the R2 indica- tor: Achieving Pareto compliance by integration,” inInternational Conference on Parallel Problem Solving from Nature. Springer, 2024, pp. 202–216

    2024

  34. [42]

    Normal-boundary intersection: A new method for generating the Pareto surface in nonlinear multicrite- ria optimization problems,

    I. Das and J. E. Dennis, “Normal-boundary intersection: A new method for generating the Pareto surface in nonlinear multicrite- ria optimization problems,”SIAM Journal on Optimization, vol. 8, no. 3, pp. 631–657, 1998

    1998

  35. [43]

    An efficient iterative approach for uniformly representing Pareto fronts,

    B. S. Saini, H. K. Singh, B. Shavazipour, and K. Miettinen, “An efficient iterative approach for uniformly representing Pareto fronts,” inInternational Conference on Evolutionary Multi-Criterion Optimization. Springer, 2025, pp. 241–256

    2025

  36. [44]

    PlatEMO: A MATLAB platform for evolutionary multi-objective optimization [educa- tional forum],

    Y. Tian, R. Cheng, X. Zhang, and Y. Jin, “PlatEMO: A MATLAB platform for evolutionary multi-objective optimization [educa- tional forum],”IEEE Computational Intelligence Magazine, vol. 12, no. 4, pp. 73–87, 2017

    2017

  37. [45]

    A repository of real-world datasets for data-driven evolutionary multiobjective optimiza- tion,

    C. He, Y. Tian, H. Wang, and Y. Jin, “A repository of real-world datasets for data-driven evolutionary multiobjective optimiza- tion,”Complex & Intelligent Systems, vol. 6, no. 1, pp. 189–197, 2020

    2020

  38. [46]

    Investigating the properties of indicators and an evolutionary many-objective algo- rithm using promising regions,

    J. Yuan, H.-L. Liu, F. Gu, Q. Zhang, and Z. He, “Investigating the properties of indicators and an evolutionary many-objective algo- rithm using promising regions,”IEEE T ransactions on Evolutionary Computation, vol. 25, no. 1, pp. 75–86, 2020

    2020

  39. [47]

    Adaptive normal vector guided evolutionary multi-and many-objective optimization,

    Y. Hua, Q. Liu, and K. Hao, “Adaptive normal vector guided evolutionary multi-and many-objective optimization,”Complex & Intelligent Systems, vol. 10, no. 3, pp. 3709–3726, 2024

    2024

  40. [48]

    A decomposition-based many-objective evolutionary algorithm updating weights when required,

    L. R. De Farias and A. F. Ara ´ujo, “A decomposition-based many-objective evolutionary algorithm updating weights when required,”swarm and evolutionary computation, vol. 68, p. 100980, 2022

    2022

  41. [49]

    A many-objective optimization evolutionary algorithm based on hyper-dominance degree,

    Z. Liu, F. Han, Q. Ling, H. Han, and J. Jiang, “A many-objective optimization evolutionary algorithm based on hyper-dominance degree,”Swarm and Evolutionary Computation, vol. 83, p. 101411, 2023

    2023

  42. [50]

    Adaptive simulated binary crossover for rotated multi-objective optimization,

    L. Pan, W. Xu, L. Li, C. He, and R. Cheng, “Adaptive simulated binary crossover for rotated multi-objective optimization,”Swarm and Evolutionary Computation, vol. 60, p. 100759, 2021

    2021

  43. [51]

    Self-adaptive polynomial muta- tion in NSGA-II,

    J. L. Carles-Bou and S. F. Gal ´an, “Self-adaptive polynomial muta- tion in NSGA-II,”Soft Computing, vol. 27, no. 23, pp. 17 711–17 727, 2023

    2023

  44. [52]

    Predefined-time distributed multiobjective optimization for network resource allocation,

    K. Zhang, L. Xu, X. Yi, Z. Ding, K. H. Johansson, T. Chai, and T. Yang, “Predefined-time distributed multiobjective optimization for network resource allocation,”Science China Information Sci- ences, vol. 66, no. 7, p. 170204, 2023

    2023

  45. [53]

    Distributed event-triggered bandit convex optimization with time-varying constraints,

    K. Zhang, X. Yi, G. Wen, M. Cao, K. H. Johansson, T. Chai, and T. Yang, “Distributed event-triggered bandit convex optimization with time-varying constraints,”IEEE T ransactions on Control of Network Systems, 2025

    2025

  46. [54]

    One- point sampling for distributed bandit convex optimization with time-varying constraints,

    K. Zhang, L. Xu, X. Yi, G. Wen, L. Xie, T. Chai, and T. Yang, “One- point sampling for distributed bandit convex optimization with time-varying constraints,”IEEE T ransactions on Control of Network Systems, 2025

    2025

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.