pith. sign in

arxiv: 2605.17454 · v1 · pith:EZGDFAHEnew · submitted 2026-05-17 · 💻 cs.AI

Multi-Party Multi-Objective Optimization as Consensus Search: Runtime Analysis of Cross-Party Recombination

Pith reviewed 2026-05-20 12:52 UTC · model grok-4.3

classification 💻 cs.AI
keywords multi-party multi-objective optimizationruntime analysiscross-party recombinationNSGA-IIconsensus searchPareto-optimal solutionsevolutionary algorithmsminimum spanning tree
0
0 comments X

The pith

Cross-party recombination in an NSGA-II variant assembles distributed templates to find common Pareto solutions in O(n log n) expected evaluations on gap benchmarks.

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

The paper treats multi-party multi-objective optimization as a consensus search task where autonomous parties maintain separate populations and must agree on solutions. It proves that a specialized CPR-NSGA-II variant reaches both common Pareto-optimal points on the MP-JCG benchmark in O(n log n) expected time by recombining complementary prefix and suffix templates held in different party groups. This stands in contrast to a payoff-guided mutation baseline that hits a gap-crossing bottleneck and requires Theta(n squared) evaluations. The work further supplies a support-cover analysis and parameterized runtime bound for the bi-party minimum spanning tree specialization, separating local filling, recombination shortcuts, and repair effects. A sympathetic reader cares because the result supplies the first runtime explanation for how recombination can shortcut the extra coverage burden that arises when decision makers cannot flatten their objectives into one shared front.

Core claim

On the MP-JCG pseudo-Boolean benchmark with an explicit gap region, a payoff-guided mutation baseline requires Theta(n squared) expected fitness evaluations to cross the gap, while an analytical CPR-NSGA-II variant discovers both common Pareto-optimal solutions in O(n log n) expected evaluations by directly assembling complementary prefix and suffix templates distributed across party populations. For the BPBOMST specialization, the symmetric average projection of each common Pareto objective vector induces an auxiliary bi-objective MST instance; suitable support representatives produce a 2 lambda common approximation cover for lambda between 1 and 2, and a representative-pool CPR-NSGA-II run

What carries the argument

Cross-party recombination (CPR) in NSGA-II, which directly combines partial solution templates maintained in separate party populations to form complete consensus solutions without extra search cost.

If this is right

  • Recombination bypasses the quadratic gap-crossing cost that mutation faces on MP-JCG.
  • Flattening the same problem into a single four-objective instance adds measurable extra coverage cost compared with the multi-party formulation.
  • The BPBOMST runtime bound isolates three distinct contributions: local auxiliary-front filling, cross-party recombination shortcuts, and edge-union repair ambiguity.
  • Each common Pareto vector reduces to an auxiliary bi-objective MST via symmetric average projection.

Where Pith is reading between the lines

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

  • If template preservation survives more realistic selection pressures, the same recombination mechanism could be tested on non-gap benchmarks to measure how much the linearithmic bound depends on the MP-JCG structure.
  • Applying the layered support-cover idea to three or more parties would show whether the approximation factor remains bounded or grows with the number of parties.
  • The separation of recombination shortcuts from local search and repair suggests that hybrid algorithms could deliberately maintain representative pools even when parties use different internal solvers.

Load-bearing premise

Party populations continue to hold and preserve the complementary prefix and suffix templates so that recombination can assemble them directly without further search.

What would settle it

An empirical run of the CPR-NSGA-II variant on MP-JCG instances that measures whether the number of fitness evaluations to reach both common solutions grows faster than O(n log n) or whether selection removes the required template diversity from the populations.

Figures

Figures reproduced from arXiv: 2605.17454 by Peilan Xu, Wenjian Luo, Xiaolei Fang.

Figure 1
Figure 1. Figure 1: Objective spaces of Party 1 (OJZJ, left) and Party 2 (G-COCZ, right) for [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Boundary-crossover mechanism on MP-JCG. Party 2 provides a near-prefix representative whose prefix contains at [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: A five-node BPBOMST instance illustrating a CPR-good shortcut. Edges [PITH_FULL_IMAGE:figures/full_fig_p032_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Fitness evaluations required to identify the common Pareto set on MP-JCG instances. Error bars represent one [PITH_FULL_IMAGE:figures/full_fig_p037_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Fitness evaluations required to achieve target approximation ratios [PITH_FULL_IMAGE:figures/full_fig_p038_5.png] view at source ↗
read the original abstract

Multi-party multi-objective optimization problems (MPMOPs) require consensus among autonomous decision makers and therefore differ from flattened many-objective formulations. Existing runtime theory for multi-objective evolutionary algorithms is largely tailored to single-party Pareto-front approximation and does not directly explain common-solution search in MPMOPs. We investigate cross-party recombination in two representative settings. On MP-JCG, a pseudo-Boolean benchmark with an explicit gap region, we prove that a payoff-guided mutation baseline faces a gap-crossing bottleneck requiring \(\Theta(n^2)\) expected fitness evaluations. In contrast, an analytical CPR-NSGA-II variant discovers both common Pareto-optimal solutions in \(O(n\log n)\) expected evaluations by directly assembling complementary prefix and suffix templates distributed across party populations. Comparing this with the flattened four-objective formulation F-JCG, our full-front coverage analysis illustrates the additional coverage burden introduced by flattening. For BPBOMST, the bi-party, two-objective-per-party specialization of the multi-party multi-objective minimum spanning tree problem, we develop a layered support-cover analysis. For each common Pareto objective vector, the symmetric average projection induces an auxiliary bi-objective MST instance, and suitable support representatives yield a \(2\lambda\)-common approximation cover with \(\lambda\in[1,2]\). We further derive an instance-parameterized expected runtime bound for a representative-pool CPR-NSGA-II variant using edge-union recombination and uniform repair. This bound separates the effects of local auxiliary-front filling, cross-party recombination shortcuts, and edge-union repair ambiguity.

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

2 major / 2 minor

Summary. The paper develops runtime theory for multi-party multi-objective optimization (MPMOPs) framed as consensus search via cross-party recombination (CPR). On the MP-JCG pseudo-Boolean benchmark it proves a Θ(n²) expected-evaluation bottleneck for payoff-guided mutation due to gap crossing, while claiming an analytical CPR-NSGA-II variant assembles both common Pareto-optimal solutions in O(n log n) expected evaluations by recombining complementary prefix/suffix templates already distributed across party populations. For the bi-party BPBOMST problem it introduces a layered support-cover analysis in which the symmetric average projection yields an auxiliary bi-objective MST instance per common Pareto vector, establishes a 2λ-common approximation cover for λ∈[1,2], and derives an instance-parameterized expected runtime bound for a representative-pool CPR-NSGA-II variant that separates local auxiliary-front filling, recombination shortcuts, and edge-union repair ambiguity. The work contrasts these results with the additional coverage burden of the flattened four-objective F-JCG formulation.

Significance. If the derivations hold, the manuscript supplies the first rigorous runtime separation between single-party Pareto approximation and multi-party consensus search, together with explicit approximation guarantees and a decomposition of runtime contributions that is rare in evolutionary computation theory. The machine-checked or fully analytical character of the MP-JCG and BPBOMST bounds, the parameter-free template-assembly shortcut, and the falsifiable prediction that CPR-NSGA-II avoids the Θ(n²) gap-crossing cost constitute genuine strengths.

major comments (2)
  1. [§4] §4 (MP-JCG runtime analysis): the O(n log n) expected-evaluation claim for the analytical CPR-NSGA-II variant rests on the unverified assumption that non-dominated sorting plus crowding distance preserves the exact complementary prefix and suffix templates in each party population at sufficient frequency to enable direct assembly without additional search cost across the gap region. No explicit invariant, drift analysis, or frequency bound is supplied to confirm that the multi-party fitness evaluations and selection dynamics maintain these building blocks; this assumption is load-bearing for the central runtime separation.
  2. [§5.2] §5.2 (BPBOMST support-cover analysis): the 2λ-common approximation cover is derived under the symmetric average projection, yet the subsequent expected-runtime bound for representative-pool CPR-NSGA-II treats the repair ambiguity of edge-union recombination as an additive O(1) factor per recombination event. It is unclear whether the λ∈[1,2] parameterization remains uniform when the auxiliary MST instances induced by different common Pareto vectors have differing edge-overlap statistics; a concrete counter-example or worst-case λ derivation would be needed to keep the bound instance-parameterized as claimed.
minor comments (2)
  1. [§5.1] Notation for the symmetric average projection is introduced without an explicit equation number; adding a numbered display would clarify its use in the layered support-cover argument.
  2. The abstract states proofs for the Θ(n²) bottleneck, O(n log n) assembly, and 2λ cover, but the main text does not cross-reference the precise lemmas or theorems that establish each part; a short proof-overview table would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful review and the positive assessment of the significance of our work. We address the major comments point by point below.

read point-by-point responses
  1. Referee: [§4] §4 (MP-JCG runtime analysis): the O(n log n) expected-evaluation claim for the analytical CPR-NSGA-II variant rests on the unverified assumption that non-dominated sorting plus crowding distance preserves the exact complementary prefix and suffix templates in each party population at sufficient frequency to enable direct assembly without additional search cost across the gap region. No explicit invariant, drift analysis, or frequency bound is supplied to confirm that the multi-party fitness evaluations and selection dynamics maintain these building blocks; this assumption is load-bearing for the central runtime separation.

    Authors: We agree that an explicit justification for the preservation of the complementary templates is necessary to fully support the O(n log n) bound. In the analytical variant presented, the multi-party fitness evaluations are designed such that individuals carrying the prefix or suffix templates receive higher selection probability, and crowding distance is applied within layers that prioritize these building blocks. However, to make this rigorous, we will add a supporting lemma providing a frequency bound and a short drift analysis showing that the templates are maintained at the required frequency throughout the optimization process. This will be incorporated in the revised manuscript. revision: yes

  2. Referee: [§5.2] §5.2 (BPBOMST support-cover analysis): the 2λ-common approximation cover is derived under the symmetric average projection, yet the subsequent expected-runtime bound for representative-pool CPR-NSGA-II treats the repair ambiguity of edge-union recombination as an additive O(1) factor per recombination event. It is unclear whether the λ∈[1,2] parameterization remains uniform when the auxiliary MST instances induced by different common Pareto vectors have differing edge-overlap statistics; a concrete counter-example or worst-case λ derivation would be needed to keep the bound instance-parameterized as claimed.

    Authors: The λ parameterization in [1,2] is derived from the properties of the symmetric average projection and holds uniformly for all common Pareto vectors, as the approximation cover depends on the worst-case overlap rather than specific statistics of individual instances. The O(1) treatment of repair ambiguity follows from the uniform repair mechanism, which resolves edge conflicts in a constant number of steps independent of the overlap. To clarify this, we will include a worst-case analysis demonstrating that the bound remains instance-parameterized without requiring per-instance adjustments to λ. We will revise the manuscript to make this explicit. revision: yes

Circularity Check

0 steps flagged

Analytical runtime derivations for CPR-NSGA-II remain self-contained without reduction to inputs by construction

full rationale

The paper's core claims consist of explicit analytical proofs and layered support-cover analyses that derive expected runtime bounds (O(n log n) for template assembly on MP-JCG; instance-parameterized bounds separating local auxiliary-front filling, cross-party recombination, and repair ambiguity on BPBOMST) directly from the benchmark structures, selection dynamics, and recombination operators. These steps are presented as structured derivations rather than fitted quantities or self-referential definitions, with the λ ∈ [1,2] range serving as an explicit approximation parameter rather than a hidden fit. No load-bearing self-citations, ansatzes smuggled via prior work, or renamings of known results are invoked to close the central arguments; the analysis is therefore independent of its own outputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard probabilistic runtime analysis techniques from evolutionary computation applied to the new multi-party consensus setting; the λ parameterization and template-assembly assumption are the main additions.

free parameters (1)
  • λ = in [1,2]
    The approximation factor in the 2λ-common approximation cover for each common Pareto objective vector in the BPBOMST analysis.
axioms (1)
  • domain assumption The symmetric average projection induces an auxiliary bi-objective MST instance for each common Pareto objective vector
    Invoked in the layered support-cover analysis to derive the approximation guarantees.

pith-pipeline@v0.9.0 · 5808 in / 1358 out tokens · 50163 ms · 2026-05-20T12:52:31.604447+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages

  1. [1]

    Bian, C., Qian, C., 2022. Better Running Time of the Non-dominated Sorting Genetic Algorithm II (NSGA-II) by Using Stochastic Tournament Selection, in: Rudolph, G., Kononova, A.V., Aguirre, H., Kerschke, P., Ochoa, G., Tušar, T. (Eds.), Parallel Problem Solving from Nature – PPSN XVII. Springer International Publishing, Cham. volume 13399, pp. 428–441

  2. [2]

    The First Proven Performance Guarantees for the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) on a Combinatorial Opti- mization Problem, in: Proc

    Cerf, S., Doerr, B., Hebras, B., Kahane, Y., Wietheger, S., 2023. The First Proven Performance Guarantees for the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) on a Combinatorial Opti- mization Problem, in: Proc. Thirty-Second Int. Jt. Conf. Artif. Intell., International Joint Conferences on Artificial Intelligence Organization, Macau, SAR China. p...

  3. [3]

    Multiparty Multiobjective Optimization By MOEA/D, in: 2022 IEEE Congr

    Chang, Y., Luo, W., Lin, X., She, Z., Shi, Y., 2022. Multiparty Multiobjective Optimization By MOEA/D, in: 2022 IEEE Congr. Evol. Comput. CEC, pp. 01–08. 39

  4. [4]

    Evolutionary Biparty Multiobjective UAV Path Planning: Problems and Empirical Comparisons

    Chen, K., Luo, W., Lin, X., Song, Z., Chang, Y., 2024. Evolutionary Biparty Multiobjective UAV Path Planning: Problems and Empirical Comparisons. IEEE Trans. Emerg. Top. Comput. Intell. 8, 2433–2445

  5. [5]

    A Novel Immune Algorithm for Multiparty Multiobjective Optimization

    Chen, K., Luo, W., Zhou, Q., Liu, Y., Xu, P., Shi, Y., 2025. A Novel Immune Algorithm for Multiparty Multiobjective Optimization. IEEE Trans. Emerg. Top. Comput. Intell. 9, 1238–1252

  6. [6]

    Three-person multi-objective conflict decision in reservoir flood control

    Chuntian, C., Chau, K., 2002. Three-person multi-objective conflict decision in reservoir flood control. European Journal of Operational Research 142, 625–631

  7. [7]

    It is difficult to tell if there is a condorcet spanning tree

    Darmann, A., 2016. It is difficult to tell if there is a condorcet spanning tree. Mathematical Methods of Operations Research 84, 93–104. doi:10.1007/s00186-016-0535-3

  8. [8]

    Finding socially best spanning trees

    Darmann, A., Klamler, C., Pferschy, U., 2011. Finding socially best spanning trees. Theory and Decision 70, 511–527

  9. [9]

    A fast and elitist multiobjective genetic algorithm: NSGA-II

    Deb, K., Pratap, A., Agarwal, S., Meyarivan, T., 2002. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Trans. Evol. Comput. 6, 182–197

  10. [10]

    Crossover can provably be useful in evolutionary computation, in: Proceedings of the 10th annual conference on Genetic and evolutionary computation, pp

    Doerr, B., Happ, E., Klein, C., 2008. Crossover can provably be useful in evolutionary computation, in: Proceedings of the 10th annual conference on Genetic and evolutionary computation, pp. 539–546

  11. [11]

    Difficulties of the NSGA-II With the Many-Objective LeadingOnes Problem

    Doerr, B., Korkotashvili, D., Krejca, M.S., 2025a. Difficulties of the NSGA-II With the Many-Objective LeadingOnes Problem. IEEE Trans. Evol. Computat. , 1–1

  12. [12]

    Doerr, B., Krejca, M.S., Opris, A., 2025b. Tight runtime guarantees from understanding the population dynamics of the GSEMO multi-objective evolutionary algorithm, in: Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25, International Joint Conferences on Artificial Intelligence Organization. pp. 8876–8884

  13. [13]

    From Understanding the Population Dynamics of the NSGA-II to the First Proven Lower Bounds

    Doerr, B., Qu, Z., 2023a. From Understanding the Population Dynamics of the NSGA-II to the First Proven Lower Bounds. AAAI 37, 12408–12416

  14. [14]

    Runtime Analysis for the NSGA-II: Provable Speed-Ups from Crossover

    Doerr, B., Qu, Z., 2023b. Runtime Analysis for the NSGA-II: Provable Speed-Ups from Crossover. AAAI 37, 12399–12407

  15. [15]

    Theoretical Analyses of Multi-Objective Evolutionary Algorithms on Multi-Modal Objectives

    Doerr, B., Zheng, W., 2021. Theoretical Analyses of Multi-Objective Evolutionary Algorithms on Multi-Modal Objectives. AAAI 35, 12293–12301

  16. [16]

    Drift analysis and average time complexity of evolutionary algorithms

    He, J., Yao, X., 2001. Drift analysis and average time complexity of evolutionary algorithms. Artificial Intelligence 127, 57–85

  17. [17]

    The analysis of evolutionary algorithms-a proof that crossover really can help

    Jansen, T., Wegener, I., 2002. The analysis of evolutionary algorithms-a proof that crossover really can help. Algorithmica 34, 47–66

  18. [18]

    Real royal road functions-where crossover provably is essential

    Jansen, T., Wegener, I., 2005. Real royal road functions-where crossover provably is essential. Discrete Applied Mathematics 149, 111–125

  19. [19]

    Running Time Analysis of Multiobjective Evolutionary Algorithms on Pseudo-Boolean Functions

    Laumanns, M., Thiele, L., Zitzler, E., 2004. Running Time Analysis of Multiobjective Evolutionary Algorithms on Pseudo-Boolean Functions. IEEE Trans. Evol. Computat. 8, 170–182

  20. [20]

    Laumanns, M., Thiele, L., Zitzler, E., Welzl, E., Deb, K., 2002. Running time analysis of multi-objective evolutionary algorithms on a simple discrete optimization problem, in: International Conference on Parallel Problem Solving from Nature, Springer. pp. 44–53

  21. [21]

    Many-objective evolutionary algorithms: A survey

    Li, B., Li, J., Tang, K., Yao, X., 2015. Many-objective evolutionary algorithms: A survey. ACM Computing Surveys (CSUR) 48, 1–35. 40

  22. [22]

    Multiagent mst cover: Pleasing all optimally via a simple voting rule, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp

    Li, B., Wu, X., Xu, C., Zhang, R., 2023. Multiagent mst cover: Pleasing all optimally via a simple voting rule, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp. 5730–5738. doi:10. 1609/aaai.v37i5.25711

  23. [23]

    Evolutionary Approach to Multiparty Multiobjective Optimization Problems with Common Pareto Optimal Solutions, in: 2020 IEEE Congr

    Liu, W., Luo, W., Lin, X., Li, M., Yang, S., 2020. Evolutionary Approach to Multiparty Multiobjective Optimization Problems with Common Pareto Optimal Solutions, in: 2020 IEEE Congr. Evol. Comput. CEC, IEEE, Glasgow, United Kingdom. pp. 1–9

  24. [24]

    Many-objective problems where crossover is provably essential

    Opris, A., 2026. Many-objective problems where crossover is provably essential. Artificial Intelligence 350, 104453

  25. [25]

    An analysis on recombination in multi-objective evolutionary optimization

    Qian, C., Yu, Y., Zhou, Z.H., 2013. An analysis on recombination in multi-objective evolutionary optimization. Artificial Intelligence 204, 99–119

  26. [26]

    A New Evolutionary Approach to Multiparty Multiobjective Optimization, in: Tan, Y., Shi, Y

    She, Z., Luo, W., Chang, Y., Lin, X., Tan, Y., 2021. A New Evolutionary Approach to Multiparty Multiobjective Optimization, in: Tan, Y., Shi, Y. (Eds.), Adv. Swarm Intell., Springer International Publishing, Cham. pp. 58–69

  27. [27]

    Multiparty distance minimization: Problems and an evolutionary approach

    She, Z., Luo, W., Lin, X., Chang, Y., Shi, Y., 2023. Multiparty distance minimization: Problems and an evolutionary approach. Swarm and Evolutionary Computation 83, 101415

  28. [28]

    An Indicator Based Evolutionary Algorithm for Multiparty Multiobjective Knapsack Problems, in: Shi, Z., Torresen, J., Yang, S

    Song, Z., Luo, W., Xu, P., Ye, Z., Chen, K., 2024. An Indicator Based Evolutionary Algorithm for Multiparty Multiobjective Knapsack Problems, in: Shi, Z., Torresen, J., Yang, S. (Eds.), Intell. Inf. Process. XII, Springer Nature Switzerland, Cham. pp. 233–246

  29. [29]

    Runtime analysis of evolutionary algorithms for multi-party multi- objective optimization

    Sun, Y., Xu, P., Luo, W., 2025. Runtime analysis of evolutionary algorithms for multi-party multi- objective optimization. IEEE Transactions on Evolutionary Computation , 1–15

  30. [30]

    Evolutionary large-scale multi-objective optimization: A survey

    Tian, Y., Si, L., Zhang, X., Cheng, R., He, C., Tan, K.C., Jin, Y., 2021. Evolutionary large-scale multi-objective optimization: A survey. ACM Comput. Surv. 54

  31. [31]

    A Negotiation-Based Multi-Objective, Multi- Party Decision-Making Model for Inter-Basin Water Transfer Scheme Optimization

    Zhang, C., Wang, G., Peng, Y., Tang, G., Liang, G., 2012. A Negotiation-Based Multi-Objective, Multi- Party Decision-Making Model for Inter-Basin Water Transfer Scheme Optimization. Water Resour Manage 26, 4029–4038

  32. [32]

    Multi-Party Multi-Objective Optimiza- tion for Discrete Problems: A Case Study on Multi-Stakeholder Recommendation

    Zhang, L., Zhu, M., Ge, Y., Yang, H., Wu, L., Zhao, H., 2025. Multi-Party Multi-Objective Optimiza- tion for Discrete Problems: A Case Study on Multi-Stakeholder Recommendation. IEEE Trans. Evol. Comput. , 1–1

  33. [33]

    MOEA/D: A Multiobjective Evolutionary Algorithm Based on Decomposition

    Zhang, Q., Li, H., 2007. MOEA/D: A Multiobjective Evolutionary Algorithm Based on Decomposition. IEEE Trans. Evol. Comput. 11, 712–731

  34. [34]

    Runtime Analysis for the NSGA-II: Proving, Quantifying, and Explaining the Inefficiency for Many Objectives

    Zheng, W., Doerr, B., 2024. Runtime Analysis for the NSGA-II: Proving, Quantifying, and Explaining the Inefficiency for Many Objectives. IEEE Trans. Evol. Computat. 28, 1442–1454

  35. [35]

    A First Mathematical Runtime Analysis of the Non-dominated Sorting Genetic Algorithm II (NSGA-II)

    Zheng, W., Liu, Y., Doerr, B., 2022. A First Mathematical Runtime Analysis of the Non-dominated Sorting Genetic Algorithm II (NSGA-II). AAAI 36, 10408–10416. 41