Pith. sign in

REVIEW 3 major objections 4 minor 48 references

Deep Reinforcement Learning-Assisted Component Auto-Configuration of Differential Evolution Algorithm for Constrained Optimization: A Foundation Model

T0 review · 3 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A deep reinforcement learning policy, trained on a diverse set of constrained problems, can automatically select per-generation mutation and constraint-handling components for Differential Evolution and generalize zero-shot to unseen proble

desk verdict SuperDE has a worthwhile idea but the empirical comparison is not interpretable as written—no budget reported, and the statistics skip SuperDE's failures. read the letter →

arxiv 2509.11016 v1 pith:ZP56YGTR submitted 2025-09-14 cs.NE

Xu Yang , Rui Wang , Kaiwen Li , Wenhua Li , Ling Wang This is my paper · ORCID
classification cs.NE
keywords deepreinforcementlearningdifferentialevolutionconstrainedoptimizationcomponentauto-configurationzero-shotgeneralizationmeta-learningDDQNconstrainthandlingtechnique
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

The paper introduces SuperDE, a deep reinforcement learning framework that automates the configuration of Differential Evolution (DE) components for constrained optimization problems (COPs). Its central claim: after offline meta-training on a diverse set of benchmark COPs, a Double Deep Q-Network policy can, at each generation, select one of four mutation strategies and one of seven constraint-handling techniques from a 22-feature population state, and this policy transfers to unseen problems in a zero-shot manner. On CEC2010, CEC2017, G2000, and BBOB2022 benchmark suites, SuperDE is reported to outperform six DE-based algorithms in finding feasible solutions and minimizing objective values or constraint violations. The paper also argues through ablations that simultaneously configuring both components yields a '1+1>2' effect, and that the offline-trained policy avoids the computational cost of per-problem online tuning.

What carries the argument

The core mechanism is an MDP that turns algorithm configuration into a sequential decision problem: state = 22 population features; action = 2D discrete tuple (mutation strategy index, CHT index) from a pool of 4 MSs (DE/rand/1, DE/best/1, DE/cur-to-best/1, DE/rand-to-best/1) and 7 CHTs (death penalty, weighted penalty, two feasibility rules, tournament, epsilon feasibility, stochastic ranking); reward = phase-dependent scalar that prioritizes violation reduction when no feasible solution exists and balances objective improvement with feasible-ratio growth otherwise. A Double Deep Q-Network with target network, experience replay, and prioritized sampling learns Q-values from trajectories col

What would settle it

A decisive test: generate a family of constrained optimization problems with a tunable parameter that changes the geometry of the feasible region (e.g., disconnectivity or the ratio of active constraints) and hold out a subset. If SuperDE's average objective/CV on the held-out instances is not better than the best fixed component pair chosen in hindsight, or if a random-configuration baseline performs equally well on problems where the oracle-optimal action lies outside the pool, then the claim that the learned policy generalizes zero-shot fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that per-generation component configuration for COPs can be learned as a Markov decision process and solved with deep reinforcement learning, producing a policy that generalizes across problem instances. The policy's input is a 22-dimensional vector of population statistics—feasibility ratio, fitness and objective extrema, constraint violation statistics, feasible-region cluster structure, correlation between objective and violation, and so on—and its output is one of 28 discrete actions, each selecting a mutation strategy and a CHT. Trained with a Double Deep Q-Network across many problem instances, this policy is applied without retraining

Load-bearing premise

The load-bearing assumption is that the 22 hand-crafted population features plus the fixed pool of 4 mutation strategies and 7 constraint-handling techniques suffice to represent the optimal configuration decision for any unseen constrained problem; if an unseen problem's best per-generation action falls outside this discretization, the zero-shot guarantee collapses (a limitation the paper acknowledges in its conclusion).

Editorial extensions

If this is right

  • If SuperDE generalizes as claimed, a single pretrained policy can serve as an off-the-shelf configurator for a broad class of constrained optimization problems, replacing per-problem hand tuning.
  • The reported advantage over SuperDE1/2/3 implies that the coordinated choice of mutation and CHT is more than the sum of independent selections; practitioners should treat these components as coupled.
  • Because training is offline, the online overhead of SuperDE is just one small network forward pass per generation, which is negligible relative to the cost of function evaluations in expensive engineering problems.
  • The paper's ablation results suggest that even random configuration (SuperDE3) is competitive on some problems, indicating that pool diversity alone contributes part of the gain; the learned policy's value is in consistency across problems.
  • The paper states that the model can be fine-tuned online; if that is effective, users could adapt the pretrained policy to a specific domain with only a small number of additional gradient updates.

Reading between the lines

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

  • One untested corollary is that the learned policy might serve as a warm start for online DRL-based configuration on a new problem class, potentially reducing the number of evaluations needed to match problem-specialized algorithms.
  • The fixed action pool is the chief limiting factor; if the state representation were combined with generative or continuous actions (e.g., F and CR values, or epsilon schedules), the same architecture could cover a much larger configuration space.
  • The 22 features include structural descriptors like feasible clusters and boundary-crossing ratio; an ablation of these features could reveal which population signals drive the policy's decisions, making the model interpretable and possibly improving state design.
  • A natural stress test would be to apply SuperDE to problems with time-varying constraints or to a combinatorial constrained problem (which the paper lists as a limitation); the zero-shot claim only covers continuous COP benchmarks.
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 / 4 minor

Summary. The paper proposes SuperDE, a Double Deep Q-Network (DDQN) controller that, at each generation of a Differential Evolution (DE) run, selects one of four mutation strategies and one of seven constraint-handling techniques. The controller is trained offline in a meta-learning fashion on CEC2010 constrained problems, and is then applied zero-shot to other benchmark suites (CEC2010 with different optima, CEC2017, G2000, BBOB2022). The authors claim that this learned per-generation component configuration significantly outperforms six DE-based baselines and also outperforms ablated variants (fixed MS, fixed CHT, random configuration). The paper includes convergence plots, Wilcoxon signed-rank test summaries, and an appendix table for BBOB2022.

Significance. If the empirical claims were fully supported, the paper would make a useful contribution to automated algorithm configuration: it demonstrates that a relatively small DDQN can be trained across a set of COPs and then used without retraining to select DE components on unseen benchmarks. The strengths are the clear MDP formulation, the inclusion of several independent held-out suites (CEC2017, G2000, BBOB2022), and an ablation study that attempts to isolate the contribution of each configured component. The authors also honestly list limitations (manual state/action design, limited training corpus, no discrete/combinatorial problems). However, the current experimental reporting is not yet sufficient to support the headline claim of zero-shot superiority: the computational budget is not specified, and the statistical significance tests are computed on a favorable subset of problems. These are load-bearing issues, not presentation details.

major comments (3)
  1. [Section IV-A and Algorithm 1 (line 21)] The evaluation setup never states the computational budget. The termination condition is left abstract as 'done←algorithmTermination(Ij)' (Algorithm 1, line 21), and no maximum number of function evaluations, population size, F, CR, or stopping rule is reported for SuperDE or for any competitor. Without resource-equivalent comparisons, the lower objective values and CVs in Tables II, IV, VI, and X could be an artifact of unequal budgets rather than evidence for the learned policy. This is central to the paper's claim of zero-shot superiority, and the setup must be specified and controlled.
  2. [Tables II and IV (Wilcoxon summaries)] The '+/-/=' counts appear to exclude comparisons in which one algorithm fails to find a feasible solution. For example, in Table II, C2010E F9: SuperDE returns NaN while CMODE, IepsilonJADE, and C2ODE find feasible solutions. This should be counted as a '+' for those competitors, but the summary row does not appear to include it. The same issue occurs in Table IV (e.g., rows where SuperDE is NaN but a competitor has a value). Excluding such cases biases the Wilcoxon test in favor of SuperDE. The authors should either include these cases as losses, or explicitly state and justify that the test is restricted to mutually feasible problems, and qualify the 'significantly outperforms' claim accordingly.
  3. [Sections III-C, III-D, III-E, and IV] The paper is not reproducible from the information provided. The training procedure lacks learning rate, optimizer, epsilon-greedy schedule, target network update frequency (Eq. (11) is not instantiated with a value), number of epochs Epo, number of instances M, and meta-learning loop details. The DE parameters F and CR for SuperDE and for all baselines are never given. Some state features are also under-specified: feature 14 relies on DBSCAN but the DBSCAN parameters are not reported, and feature 15 ('ratio of feasible boundary crossings') is not defined. No source code is provided. Given that the contribution is exclusively empirical, these omissions are major.
minor comments (4)
  1. [Section IV-A] The definition of 'C2010E' is vague: 'with different optima' is not explained. Please clarify how the optima are shifted and whether this creates genuinely different instances from the CEC2010 training set.
  2. [Section III-C.1] Features 14 and 15 lack precise definitions. For DBSCAN-based clustering, the radius and minimum samples parameters should be stated; the computation of 'feasible boundary crossings' should be formalized.
  3. [Section III-C.2] The candidate pool is small (4 MSs and 7 CHTs) and fixed. The authors acknowledge this limitation in Section V, but the abstract's 'foundation model' terminology seems disproportionate for a policy with a 28-action discrete space trained on one benchmark suite. A more measured term such as 'pretrained controller' would be more accurate.
  4. [Tables and figures] Some table entries have no +/-/= marker when the value is NaN (e.g., Table II), which makes the summary rows difficult to verify. Figure 4 is hard to read in the text; the legend and line colors should be improved. References [24] and [34] appear to be duplicates of the same Kiran et al. paper and should be merged.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the policy is trained on CEC2010 with a per-generation reward and evaluated on held-out suites (CEC2017, G2000, BBOB2022) against external baselines; no load-bearing step reduces to its own inputs.

full rationale

The derivation chain is self-contained. The DDQN policy is trained offline on CEC2010 instances using the per-generation reward in Eqs. (8)-(9), and the generalization claims are tested on genuinely held-out suites in RQ2 (CEC2017, G2000, BBOB2022), where the paper reports objective/CV values and Wilcoxon tests against six external DE variants. RQ1's CEC2010-extension test is explicitly disclosed as a variant of the training suite, so it is not presented as evidence of zero-shot generalization. The reward is a per-generation proxy aligned with, but not identical to, the final evaluation metric; training to maximize one and then reporting the other is standard RL methodology, not a definitional reduction. The manually defined 22-feature state space and 28-action pool are clearly stated design choices (Section III-C/E), not smuggled in via self-citation. The self-citations in the related work ([10], [14], [15], and [40]) are contextual and not load-bearing for any derived claim. Section V honestly acknowledges limitations: the manual state/action design lacks flexibility and the training corpus is limited by dataset size; these weaken the breadth of the zero-shot claim but do not make the derivation circular. Finally, the absence of a reported computational budget is a fairness/reproducibility concern, not a circularity concern, and does not affect this score.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The method rests on hand-designed state/action spaces, reward alignment, and transfer from an 18-function training suite. DE parameters F/CR and most DRL hyperparameters are unreported, so they behave as free parameters in any replication attempt. No new physical or conceptual entities are introduced.

free parameters (7)
  • Reward weights in Eq. (8): 0.4 fitness, 0.6 fsr = 0.4/0.6
    Hand-chosen weights balancing best-fitness improvement against feasible ratio; no sensitivity analysis is provided.
  • Reward weights in Eq. (9): 0.7 avgCV, 0.3 minCV = 0.7/0.3
    Hand-chosen weights for infeasible populations; no sensitivity analysis is provided.
  • CHT2 penalty coefficient cp = 0.3
    Fixed penalty coefficient in the weighted penalty constraint handler (Eq. 17).
  • CHT6 epsilon threshold = 0.05
    Relaxation threshold for epsilon-feasibility ranking; fixed hand-set value.
  • CHT7 stochastic ranking probability Pf = 0.45
    Probability of comparing by objective in stochastic ranking; standard from Runarsson and Yao but a hand-set constant.
  • DE parameters F and CR = not reported
    The mutation factor and crossover rate are never specified; if fixed, they are free parameters, if adaptive, the adaptation is undisclosed.
  • DDQN hyperparameters = batch 128, buffer 20000, hidden 256; others unstated
    Learning rate, discount gamma, target update frequency, epsilon schedule, epoch count, and instance count are omitted, making the exact training procedure underdetermined.
assumptions (5)
  • domain assumption The 22 population-state features are a sufficient statistic for optimal component selection.
    Section III-C.1 lists 22 features; if a discriminative property is missing, the policy cannot choose correctly.
  • domain assumption The pool of 4 mutation strategies and 7 CHTs contains a near-optimal component for every stage of every test problem.
    Section III-E defines the fixed pool; the zero-shot claim assumes coverage of unseen problems.
  • domain assumption Reward functions (Eq. 8 and 9) are aligned with final solution quality and feasibility.
    Section III-C.3 defines rewards; if maximizing them does not improve end-of-run best objective, the policy is not useful.
  • domain assumption Training on the 18 CEC2010 functions transfers to CEC2017, G2000, and BBOB2022.
    Section IV-A states CEC2010 is the training suite; Section IV-D claims transfer; Section V admits the corpus is small, weakening this assumption.
  • standard math Standard MDP and RL convergence assumptions hold for the evolutionary process.
    Section II-A assumes an MDP with fixed transition probabilities, though the EA's transitions are stochastic and nonstationary in practice.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Reinforcement Learning-Assisted Component Auto-Configuration of Differential Evolution Algorithm for Constrained Optimization: A Foundation Model." pith.science (2026). https://pith.science/paper/ZP56YGTR

@misc{pith2026250911016,
  author       = {Pith},
  title        = {Pith review of: Deep Reinforcement Learning-Assisted Component Auto-Configuration of Differential Evolution Algorithm for Constrained Optimization: A Foundation Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZP56YGTR}},
  note         = {Machine review of arXiv:2509.11016}
}
read the original abstract

Despite significant efforts to manually design high-performance evolutionary algorithms, their adaptability remains limited due to the dynamic and ever-evolving nature of real-world problems. The "no free lunch" theorem highlights that no single algorithm performs optimally across all problems. While online adaptation methods have been proposed, they often suffer from inefficiency, weak convergence, and limited generalization on constrained optimization problems (COPs). To address these challenges, we introduce a novel framework for automated component configuration in Differential Evolution (DE) algorithm to address COPs, powered by Deep Reinforcement Learning (DRL). Specifically, we propose SuperDE, a foundation model that dynamically configures DE's evolutionary components based on real-time evolution. Trained offline through meta-learning across a wide variety of COPs, SuperDE is capable of recommending optimal per-generation configurations for unseen problems in a zero-shot manner. Utilizing a Double Deep Q-Network (DDQN), SuperDE adapts its configuration strategies in response to the evolving population states during optimization. Experimental results demonstrate that SuperDE significantly outperforms existing state-of-the-art algorithms on benchmark test suites, achieving superior generalization and optimization performance.

Figures

Figures reproduced from arXiv: 2509.11016 by the authors.

Figure 1
Figure 1. Overall structure of SuperDE [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The schematic diagrams of traditional adaptive DE, online DRL-based adaptive DE, and our proposed foundation model SuperDE. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The workflow of DDQN where w denotes the sample weights (from prioritized experience replay). The target network parameters θ − are periodically syn￾chronized with the critic network to propagate learned features. This is done either every target update frequency steps according to Eq. (11): θ − ← (1 − ξ) · θ − + ξ · θ (11) where ξ is the target smooth factor, balancing stability (small ξ) and adaptability (large ξ)… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: CV convergence of different algorithms on CEC2010 extension F4, F5, F6, F10, F11, and F12. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references

  1. [1]

    Engineering design: a sys- tematic approach,

    W. Beitz, G. Pahl, and K. Grote, “Engineering design: a sys- tematic approach,” Mrs Bulletin , vol. 71, p. 30, 1996

    1996

  2. [2]

    A knowledge- guided bi-population evolutionary algorithm for energy-efficient scheduling of distributed flexible job shop problem,

    F. Yu, C. Lu, J. Zhou, L. Yin, and K. Wang, “A knowledge- guided bi-population evolutionary algorithm for energy-efficient scheduling of distributed flexible job shop problem,” Engineer- ing Applications of Artificial Intelligence , vol. 128, p. 107458, 2024

    2024

  3. [3]

    Path planning techniques for mobile robots: Review and prospect,

    L. Liu, X. Wang, X. Yang, H. Liu, J. Li, and P. Wang, “Path planning techniques for mobile robots: Review and prospect,” Expert Systems with Applications, vol. 227, p. 120254, 2023

    2023

  4. [4]

    Review of ship arrangement design using optimization methods,

    K.-S. Kim and M.-I. Roh, “Review of ship arrangement design using optimization methods,” Journal of Computational Design and Engineering, vol. 12, no. 1, pp. 100–121, 2025

    2025

  5. [5]

    A brief review of portfolio optimization techniques,

    A. Gunjan and S. Bhattacharyya, “A brief review of portfolio optimization techniques,” Artificial Intelligence Review, vol. 56, no. 5, pp. 3847–3886, 2023

    2023

  6. [6]

    An adaptive scheduling system with genetic algorithms for arranging employee training programs,

    Y.-S. Juang, S.-S. Lin, and H.-P. Kao, “An adaptive scheduling system with genetic algorithms for arranging employee training programs,” Expert Systems with Applications, vol. 33, no. 3, pp. 642–651, 2007

    2007

  7. [7]

    Constraint-handling techniques used with evolutionary algorithms,

    C. A. C. Coello, “Constraint-handling techniques used with evolutionary algorithms,” in Proceedings of the genetic and evo- lutionary computation conference companion , 2022, pp. 1310– 1333

    2022

  8. [8]

    A review on constraint handling techniques for population- based algorithms: from single-objective to multi-objective opti- mization,

    I. Rahimi, A. H. Gandomi, F. Chen, and E. Mezura-Montes, “A review on constraint handling techniques for population- based algorithms: from single-objective to multi-objective opti- mization,” Archives of Computational Methods in Engineering , vol. 30, no. 3, pp. 2181–2209, 2023

    2023

Show all 48 references
  1. [9]

    A survey on evolutionary constrained multiobjective optimization,

    J. Liang, X. Ban, K. Yu, B. Qu, K. Qiao, C. Yue, K. Chen, and K. C. Tan, “A survey on evolutionary constrained multiobjective optimization,” IEEE Transactions on Evolutionary Computa- tion, vol. 27, no. 2, pp. 201–221, 2022

    2022

  2. [10]

    Two-archive evolutionary algorithm for constrained multiobjective optimization,

    K. Li, R. Chen, G. Fu, and X. Yao, “Two-archive evolutionary algorithm for constrained multiobjective optimization,” IEEE Transactions on Evolutionary Computation , vol. 23, no. 2, pp. 303–315, 2018

    2018

  3. [11]

    A novel differential evolution algorithm for solving constrained engineering optimization problems,

    A. W. Mohamed, “A novel differential evolution algorithm for solving constrained engineering optimization problems,” Jour- nal of Intelligent Manufacturing , vol. 29, pp. 659–692, 2018

    2018

  4. [12]

    A constrained multiobjective evolutionary algorithm with detect-and-escape strategy,

    Q. Zhu, Q. Zhang, and Q. Lin, “A constrained multiobjective evolutionary algorithm with detect-and-escape strategy,” IEEE Transactions on Evolutionary Computation , vol. 24, no. 5, pp. 938–947, 2020

    2020

  5. [13]

    A random forest-assisted evolutionary algorithm for data-driven constrained multiobjective combina- torial optimization of trauma systems,

    H. Wang and Y. Jin, “A random forest-assisted evolutionary algorithm for data-driven constrained multiobjective combina- torial optimization of trauma systems,” IEEE transactions on cybernetics, vol. 50, no. 2, pp. 536–549, 2018

    2018

  6. [14]

    A dual-population-based evolutionary algorithm for constrained multiobjective optimization,

    M. Ming, A. Trivedi, R. Wang, D. Srinivasan, and T. Zhang, “A dual-population-based evolutionary algorithm for constrained multiobjective optimization,” IEEE Transactions on Evolution- ary Computation, vol. 25, no. 4, pp. 739–753, 2021

    2021

  7. [15]

    Constrained multi- objective optimization with deep reinforcement learning assisted operator selection,

    F. Ming, W. Gong, L. Wang, and Y. Jin, “Constrained multi- objective optimization with deep reinforcement learning assisted operator selection,” IEEE/CAA Journal of Automatica Sinica , vol. 11, no. 4, pp. 919–931, 2024

    2024

  8. [16]

    Ensemble strate- gies for population-based optimization algorithms–a survey,

    G. Wu, R. Mallipeddi, and P. N. Suganthan, “Ensemble strate- gies for population-based optimization algorithms–a survey,” Swarm and evolutionary computation , vol. 44, pp. 695–711, 2019

    2019

  9. [17]

    Adaboost-inspired multi-operator ensemble strategy for multi-objective evolution- ary algorithms,

    C. Wang, R. Xu, J. Qiu, and X. Zhang, “Adaboost-inspired multi-operator ensemble strategy for multi-objective evolution- ary algorithms,” Neurocomputing, vol. 384, pp. 243–255, 2020

    2020

  10. [18]

    Self-adaptive multi-objective evolutionary algorithm for flexible job shop scheduling with fuzzy processing time,

    R. Li, W. Gong, and C. Lu, “Self-adaptive multi-objective evolutionary algorithm for flexible job shop scheduling with fuzzy processing time,” Computers & Industrial Engineering , vol. 168, p. 108099, 2022

    2022

  11. [19]

    A multi-objective evolutionary algorithm with interval based initialization and self-adaptive crossover operator for large-scale feature selection in classifica- tion,

    Y. Xue, X. Cai, and F. Neri, “A multi-objective evolutionary algorithm with interval based initialization and self-adaptive crossover operator for large-scale feature selection in classifica- tion,” Applied Soft Computing, vol. 127, p. 109420, 2022

    2022

  12. [20]

    An adap- tive differential evolution algorithm based on belief space and generalized opposition-based learning for resource allocation,

    W. Deng, H. Ni, Y. Liu, H. Chen, and H. Zhao, “An adap- tive differential evolution algorithm based on belief space and generalized opposition-based learning for resource allocation,” Applied Soft Computing, vol. 127, p. 109419, 2022

    2022

  13. [21]

    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 Transactions JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2025 14 on Emerging Topics in Computati...

    2025

  14. [22]

    Deep reinforcement learning-guided coevolutionary algorithm for constrained mul- tiobjective optimization,

    W. Luo, X. Yu, G. G. Yen, and Y. Wei, “Deep reinforcement learning-guided coevolutionary algorithm for constrained mul- tiobjective optimization,” Information Sciences , vol. 692, p. 121648, 2025

    2025

  15. [23]

    Deep reinforcement learning: A brief survey,

    K. Arulkumaran, M. P. Deisenroth, M. Brundage, and A. A. Bharath, “Deep reinforcement learning: A brief survey,” IEEE Signal Processing Magazine, vol. 34, no. 6, pp. 26–38, 2017

    2017

  16. [24]

    Deep reinforcement learning for au- tonomous driving: A survey,

    B. R. Kiran, I. Sobh, V. Talpaert, P. Mannion, A. A. A. Sallab, S. Yogamani, and P. P´ erez, “Deep reinforcement learning for au- tonomous driving: A survey,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 6, pp. 4909–4926, 2022

    2022

  17. [25]

    Bi-phase episodic memory-guided deep reinforcement learning for robot skills,

    D. Liu, Y. Jiang, Y. Du, Z. Wang, and M. Cong, “Bi-phase episodic memory-guided deep reinforcement learning for robot skills,” IEEE Robotics and Automation Letters , vol. 10, no. 9, pp. 8674–8681, 2025

    2025

  18. [26]

    Applications of deep reinforcement learning in communications and networking: A survey,

    N. C. Luong, D. T. Hoang, S. Gong, D. Niyato, P. Wang, Y.- C. Liang, and D. I. Kim, “Applications of deep reinforcement learning in communications and networking: A survey,” IEEE Communications Surveys & Tutorials, vol. 21, no. 4, pp. 3133– 3174, 2019

    2019

  19. [27]

    Reconfigurable intelligent surfaces for wire- less communications: Principles, challenges, and opportunities,

    M. A. ElMossallamy, H. Zhang, L. Song, K. G. Seddik, Z. Han, and G. Y. Li, “Reconfigurable intelligent surfaces for wire- less communications: Principles, challenges, and opportunities,” IEEE Transactions on Cognitive Communications and Net- working, vol. 6, no. 3, pp. 990–1002, 2020

    2020

  20. [28]

    A self-adaptive evolutionary multi-task based con- strained multi-objective evolutionary algorithm,

    K. Qiao, J. Liang, K. Yu, M. Wang, B. Qu, C. Yue, and Y. Guo, “A self-adaptive evolutionary multi-task based con- strained multi-objective evolutionary algorithm,” IEEE Trans- actions on Emerging Topics in Computational Intelligence , vol. 7, no. 4, pp. 1098–1112, 2023

    2023

  21. [29]

    A constrained multiobjective evolutionary algorithm based on adaptive con- straint regulation,

    F. Gu, H. Liu, Y.-m. Cheung, and H.-L. Liu, “A constrained multiobjective evolutionary algorithm based on adaptive con- straint regulation,” Knowledge-Based Systems , vol. 260, p. 110112, 2023

    2023

  22. [30]

    A dual- population evolutionary algorithm based on adaptive constraint strength for constrained multi-objective optimization,

    K. Yang, J. Zheng, J. Zou, F. Yu, and S. Yang, “A dual- population evolutionary algorithm based on adaptive constraint strength for constrained multi-objective optimization,” Swarm and Evolutionary Computation , vol. 77, p. 101247, 2023

    2023

  23. [31]

    Aschea: new results using adaptive segregational constraint handling,

    S. B. Hamida and M. Schoenauer, “Aschea: new results using adaptive segregational constraint handling,” in Proceedings of the 2002 Congress on Evolutionary Computation. CEC’02 (Cat. No. 02TH8600), vol. 1. IEEE, 2002, pp. 884–889

    2002

  24. [32]

    Dynamic neighborhood hybrid par- ticle swarm optimization for constrained optimization,

    P. Hu and D. Chang-shou, “Dynamic neighborhood hybrid par- ticle swarm optimization for constrained optimization,” in 2010 International Conference on Computational and Information Sciences, 2010

    2010

  25. [33]

    Efficient constrained optimization by the ε constrained adaptive differential evolution,

    T. Takahama and S. Sakai, “Efficient constrained optimization by the ε constrained adaptive differential evolution,” in IEEE congress on evolutionary computation. IEEE, 2010, pp. 1–8

    2010

  26. [34]

    Deep reinforcement learning for au- tonomous driving: A survey,

    B. R. Kiran, I. Sobh, V. Talpaert, P. Mannion, A. A. Al Sallab, S. Yogamani, and P. P´ erez, “Deep reinforcement learning for au- tonomous driving: A survey,” IEEE transactions on intelligent transportation systems, vol. 23, no. 6, pp. 4909–4926, 2021

    2021

  27. [35]

    Deep reinforcement learning control for radar detection and tracking in congested spectral environ- ments,

    C. E. Thornton, M. A. Kozy, R. M. Buehrer, A. F. Martone, and K. D. Sherbondy, “Deep reinforcement learning control for radar detection and tracking in congested spectral environ- ments,” IEEE Transactions on Cognitive Communications and Networking, vol. 6, no. 4, pp. 1335–1349, 2020

    2020

  28. [36]

    Deep reinforcement learning based mobile robot navigation: A review,

    K. Zhu and T. Zhang, “Deep reinforcement learning based mobile robot navigation: A review,”Tsinghua Science and Tech- nology, vol. 26, no. 5, pp. 674–691, 2021

    2021

  29. [37]

    Adaptive constraint handling technique selection for constrained multi-objective op- timization,

    C. Wang, Z. Liu, J. Qiu, and L. Zhang, “Adaptive constraint handling technique selection for constrained multi-objective op- timization,” Swarm and Evolutionary Computation , vol. 86, p. 101488, 2024

    2024

  30. [38]

    Principled design of translation, scale, and rotation invariant variation operators for metaheuristics,

    Y. Tian, X. Zhang, C. He, K. C. Tan, and Y. Jin, “Principled design of translation, scale, and rotation invariant variation operators for metaheuristics,” Chinese Journal of Electronics , vol. 32, no. 1, pp. 111–129, 2023

    2023

  31. [39]

    Problem definitions and evaluation criteria for the cec 2010 competition on constrained real-parameter optimization,

    R. Mallipeddi and P. N. Suganthan, “Problem definitions and evaluation criteria for the cec 2010 competition on constrained real-parameter optimization,” in IEEE Congress on Evolution- ary Computation, Technical Report, vol. 24, 2010, p. 910

    2010

  32. [40]

    Problem defini- tions and evaluation criteria for the cec 2017 competition on constrained real-parameter optimization,

    G. Wu, R. Mallipeddi, and P. N. Suganthan, “Problem defini- tions and evaluation criteria for the cec 2017 competition on constrained real-parameter optimization,” in IEEE Congress on Evolutionary Computation, Technical Report, vol. 9. National University of Defense Technology...

    2017

  33. [41]

    Stochastic ranking for constrained evolutionary optimization,

    T. P. Runarsson and X. Yao, “Stochastic ranking for constrained evolutionary optimization,” IEEE Transactions on evolutionary computation, vol. 4, no. 3, pp. 284–294, 2000

    2000

  34. [42]

    Building scalable test problems for benchmark- ing constrained optimizers,

    P. Dufoss´ e, N. Hansen, D. Brockhoff, P. R. Sampaio, A. Atamna, and A. Auger, “Building scalable test problems for benchmark- ing constrained optimizers,” Technical Report. http://numbbo. github. io/coco-doc/bbob-constrained/To be submitted to the SIAM Journal of Optimization...

    2022

  35. [43]

    Differential evolution–a simple and effi- cient heuristic for global optimization over continuous spaces,

    R. Storn and K. Price, “Differential evolution–a simple and effi- cient heuristic for global optimization over continuous spaces,” Journal of global optimization, vol. 11, no. 4, pp. 341–359, 1997

    1997

  36. [44]

    Combining multiobjective optimization with differential evolution to solve constrained optimization problems,

    Y. Wang and Z. Cai, “Combining multiobjective optimization with differential evolution to solve constrained optimization problems,” IEEE Transactions on Evolutionary Computation , vol. 16, no. 1, pp. 117–134, 2012

    2012

  37. [45]

    Adaptive differential evolution with multi-population-based mutation operators for constrained optimization,

    B. Xu, L. Tao, X. Chen, and W. Cheng, “Adaptive differential evolution with multi-population-based mutation operators for constrained optimization,” Soft Computing, vol. 23, no. 10, pp. 3423–3447, 2019

    2019

  38. [46]

    An enhanced adaptive differential evolution approach for constrained opti- mization problems

    W. Yi, Z. Lin, Y. Chen, Z. Pei, and J. Lu, “An enhanced adaptive differential evolution approach for constrained opti- mization problems. ”CMES-Computer Modeling in Engineering & Sciences, vol. 136, no. 3, 2023

    2023

  39. [47]

    Composite dif- ferential evolution for constrained evolutionary optimization,

    B.-C. Wang, H.-X. Li, J.-P. Li, and Y. Wang, “Composite dif- ferential evolution for constrained evolutionary optimization,” IEEE Transactions on Systems, Man, and Cybernetics: Sys- tems, vol. 49, no. 7, pp. 1482–1495, 2018

    2018

  40. [48]

    A multiobjective optimization-based evolutionary algorithm for constrained optimization,

    Z. Cai and Y. Wang, “A multiobjective optimization-based evolutionary algorithm for constrained optimization,” IEEE Transactions on evolutionary computation , vol. 10, no. 6, pp. 658–675, 2006. Appendix Table X present experimental results on 10D BBOB2022 for RQ2. The experime...

    2006

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.