Pith. sign in

REVIEW 4 major objections 5 minor 33 references

Dual-Individual Genetic Algorithm: A Dual-Individual Approach for Efficient Training of Multi-Layer Neural Networks

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A two-individual genetic algorithm with a Leader and a Follower can train small cat-versus-non-cat networks to match or beat gradient descent on test accuracy for shallow architectures, while overfitting on deeper ones.

desk verdict The leader-follower GA variant is coherent and honestly described, but the headline accuracy gap is an artifact of picking the best test result from a ten-architecture Pareto set on a 50-image test set. read the letter →

arxiv 2504.17346 v2 pith:KBYQOME7 submitted 2025-04-24 cs.NE cs.AI

Tran Thuy Nga Truong , Jooyong Kim This is my paper · ORCID
classification cs.NEcs.AI
keywords GeneticAlgorithmdual-individualevolutionLeader-FolloweroptimizationneuralnetworktrainingbinaryimageclassificationParetodominanceself-adaptivelayerdimensionsGaussianmutation
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 proposes a genetic algorithm that keeps exactly two evolving candidate solutions, labeled Leader and Follower, and uses their recombination and mutation to train small feed-forward neural networks for binary image classification. The Leader is the current best-cost solution and drives exploitation, while the Follower preserves diversity and can replace the Leader when it improves, so the two roles swap throughout optimization. The method also evolves hidden-layer widths inside fixed bounds, removing manual architecture choice. On the paper's cat-versus-non-cat benchmark, the strongest reported result is a three-layer network reaching 99.04% training and 80% test accuracy with cost 0.06, compared with 99.52% training and 74% test accuracy for gradient descent at cost 0.0378; on a four-layer network the method overfits and falls to 76% test accuracy versus 80% for gradient descent. A sympathetic reading is that evolutionary search with two individuals can match or beat gradient descent on shallow networks while carrying an overfitting risk on deeper ones.

What carries the argument

The load-bearing object is the Leader-Follower pair: two parameter dictionaries for weights and biases plus two small lists of candidate layer architectures, with the roles assigned by Pareto-dominance rank and cost, and swapped when the Follower's best solution beats the Leader's. Each generation performs row-wise multi-point crossover on weight matrices, arithmetic crossover for biases, Gaussian mutation with a mutation rate that decays from 0.9 to 0.1 over 20,000 iterations, and architecture generation that resembles Harmony Search with a 0.9 consideration rate and a 0.3 pitch-adjustment rate, weighted by cost through roulette-wheel selection. An anabolism step merges leader, follower, and offspring solutions, trims parameter matrices to the chosen layer widths, and fills any gaps with fresh architectures so both agents keep five candidates. This machinery is what lets the method search weights and architectures simultaneously with only two persistent individuals.

What would settle it

Re-run the [12288, 17, 4, 1] comparison many times with fresh random seeds and new 209/50 train/test splits and compare mean test accuracy; since the reported 80% versus 74% is just 40 versus 37 correct images out of 50, the claim stands only if the evolutionary method's mean test accuracy stays above gradient descent's across those runs.

Watch

Extended reading notes

Core claim

The central discovery the authors claim is that a population of two role-labeled neural-network parameter sets is enough to train a multilayer perceptron: crossover between Leader and Follower, with Gaussian mutation and Pareto-cost ranking of alternative layer architectures, produces weights that classify a small image set as accurately as gradient descent on two- and three-layer networks, and sometimes more accurately on held-out data. The headline numbers are a three-layer network [12288, 17, 4, 1] with 99.04% training and 80% test accuracy versus the same architecture trained by gradient descent at 99.52% training and 74% test, and a two-layer network [12288, 7, 1] at 100% and 74% versus 100% and 72%. The authors also report that on a four-layer network [12288, 20, 7, 5, 1] the genetic method reaches 99.52% training accuracy but only 76% test accuracy, below gradient descent's 98% and 80%, which they attribute to overfitting from aggressive joint optimization of weights and architecture. They interpret the layer-dimension results as evidence that the algorithm can refine hidden-layer sizes on its own, because the evolved configurations settle onto the same values as the optimized final architectures.

Load-bearing premise

The comparison rests on a single random seed and a single train/test split of 209 training and 50 test images, so the reported 6-point test-accuracy edge could be noise rather than a systematic property of the method.

Editorial extensions

If this is right

  • If the central claim holds, evolutionary training with two individuals can replace manual architecture search for small shallow classifiers: the same run that optimizes weights also picks hidden-layer sizes within the declared maximum.
  • On the three-layer benchmark the genetic method's test accuracy is 80% versus 74% for gradient descent on the same architecture, implying a 6-point generalization gain on a 50-image test set.
  • The paper's own four-layer result is worse on test accuracy (76% versus 80%) despite higher training accuracy, so the claim of advantage is limited to shallow networks unless regularization is added.
  • The reported cost values, 0.06 for the genetic three-layer solution versus 0.0378 for gradient descent, show the evolutionary solution reaches comparable or better test accuracy at a higher training cost.
  • The dynamic role-swapping between Leader and Follower, with the Follower allowed to replace the Leader, is presented as the mechanism that balances exploitation and exploration without eliminating either candidate.

Reading between the lines

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

  • Editorial inference: a repeated-seed study is the natural next test; with only 50 test images, the 80%-versus-74% gap is three images, and the paper itself notes run-to-run variance from random mutations, so the reported edge could shrink or reverse under repeated runs.
  • Editorial inference: the role-swapping design suggests a cheap ensemble, namely averaging the Leader and Follower predictions at the end of training, since the two agents are maintained as distinct solutions and may capture different regions of the loss surface; this is not tested in the paper.
  • Editorial inference: the mutation schedule, Gaussian noise with fixed scale 0.008 and a rate decaying from 0.9 to 0.1, may alone account for much of the exploration; comparing the dual-individual design against a single-individual evolution strategy with the same mutation budget would isolate the contribution of the Leader-Follower roles.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. This paper proposes a two-individual (Leader-Follower) genetic algorithm for training multilayer perceptrons, with self-adaptive hidden-layer dimensions. The method evolves two parameter sets through row-based crossover, Gaussian mutation, and Pareto-rank-based architecture selection, and is evaluated on the 209-image/50-image cat-vs-non-cat dataset from the Deep Learning Specialization. The paper claims that the proposed method outperforms gradient descent on two-layer and three-layer networks and discusses overfitting in the four-layer case.

Significance. The algorithmic description is unusually detailed, and the authors candidly acknowledge overfitting and run-to-run variability; the proposed mechanism is a reasonable incremental variant of (1+1)-ES combined with harmony-search-like architecture generation. However, the central empirical claim that the evolutionary method beats gradient descent is not established by the evidence in Table 2: the reported 80% test accuracy is the best test accuracy among ten Pareto solutions, obtained from a single seed on a 50-image test set. These issues are load-bearing for the main claim and require new experiments and a pre-specified model-selection rule to resolve.

major comments (4)
  1. [Section 3, Table 2] The reported [12288,17,4,1] result with 80% test accuracy is not the solution selected by Algorithm 1. Algorithm 1 (line 17) returns the minimum-cost leader solution; for max_layer_dims=[12288,20,5,1] that solution is [12288,20,3,1] (cost 0.03391, test 78%), not [12288,17,4,1] (cost 0.05926, test 80%). Because the paper never specifies a validation-based selection rule, the 80% figure appears to be the maximum test accuracy over the ten Pareto solutions, chosen after observing the 50-image test set. This selection bias inflates the reported accuracy and invalidates the head-to-head comparison with gradient descent. Please report the test accuracy of the solution actually returned by the algorithm, or use a pre-registered validation selection rule, and show that the conclusion is unchanged.
  2. [Section 3, Table 2] The comparison rests on a single run with random.seed(42) and a 50-image test set, where 80% versus 74% is a difference of three images. The paper itself states that random mutations cause high run-to-run variability. Provide multiple seeds and, ideally, multiple train/test splits, with means and standard deviations (or confidence intervals) and a paired significance test; otherwise the 6-point test gap is within the noise of the experimental setup.
  3. [Abstract and Section 3] The abstract's headline comparison is apples-to-oranges: it contrasts a three-layer proposed network (80% test) with a four-layer gradient-descent network (80% test), while Section 3 claims superiority over gradient descent on the same architecture [12288,17,4,1] (80% vs 74%). These two claims use different comparison protocols, and the abstract's phrase 'higher performance' is not supported when the test accuracies are equal. Please unify the comparison protocol and ensure that every superiority claim uses the same architecture for both methods.
  4. [Sections 2.3 and 2.7, Algorithm 1] The mutation scale is set to 0.0008 in Algorithm 1 (line 8) but described as 0.008 in the text and in Figure 4. This discrepancy must be resolved, since mutation_scale is a free parameter of the method and directly affects the reported results.
minor comments (5)
  1. [Section 3, first paragraph] The sentence 'the propsed method outperforms gradient' is inaccurate as stated, because for [12288,17,4,1] the proposed method's training accuracy (99.04%) is lower than gradient descent's (99.52%); the claim should be limited to test accuracy.
  2. [Section 2.5, Function 1] CR, PAR, and the mutation parameters are fixed ad hoc values, and no sensitivity analysis is reported. At minimum, state how these hyperparameters were chosen and whether the qualitative comparison is robust to reasonable variations.
  3. [Introduction, Reference [17]] Reference [17] is the paper's own arXiv entry, cited as the source for the (1+1)-ES approach it differs from. This is circular; a standard (1+1)-ES reference should be used instead.
  4. [Data Availability Statement] The statement that source code is 'available upon request' limits reproducibility; please deposit the Jupyter Notebook in a permanent public repository with a DOI.
  5. [Table 2 and Figure 4] The table entries mix architecture vectors with cost values inside the same bracket notation, and Figure 4's mutation-rate curve should specify whether the rate starts at 0.9 and decays to 0.1 or vice versa; the current text and Algorithm 1 are ambiguous about the direction of the decay.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the GA results are empirical benchmarks, not predictions derived from fitted inputs; the only self-citation is not load-bearing.

full rationale

The paper makes no first-principles derivation that could reduce to its inputs. Section 2 specifies a Genetic Algorithm that minimizes the cross-entropy training cost (Eq. 2) through mutation, crossover, and Pareto-based selection; Section 3 reports the resulting train/test accuracies as empirical measurements. None of the reported accuracies is obtained by substituting a fitted parameter into a formula that defines the same quantity, and Algorithm 1's best-cost output is not used as a hidden input to a later prediction. The single self-citation, ref [17], is the authors' own arXiv posting and appears in the introduction when contrasting the proposed two-individual scheme with (1+1)-ES; it is not used to justify any convergence, uniqueness, or performance claim, so it is not load-bearing. The concern that Table 2 reports the best test accuracy among the ten Pareto solutions rather than the minimum-cost solution is a statistical validity and fair-comparison issue, not circularity as defined here: the reported number is still a measured outcome, not a quantity equivalent by construction to an input. The paper itself acknowledges overfitting and run-to-run variance, which are limitations rather than signs of circular derivation. On the requested circularity definition, the empirical derivation chain is self-contained and the comparison, however contestable statistically, is not circular.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical entities. It relies on a large set of hand-chosen hyperparameters and ad hoc algorithmic choices, such as zero initialization, the 1/(1+cost) fitness, even-odd assignment, and the custom Pareto rank. The 'self-adaptive architecture' claim is weakened by the manual max_layer_dims bounds, which still require user input.

free parameters (8)
  • CR (Consideration Rate) = 0.9
    Hand-selected probability of reusing a hidden-layer value from existing solutions in create_new_solution (Function 1).
  • PAR (Pitch Adjustment Rate) = 0.3
    Hand-selected probability of perturbing a selected layer size by -2 to 2 in Function 1.
  • mutation_scale = 0.0008 (Algorithm 1) vs 0.008 (Section 2.7 and Figure 4)
    Gaussian noise scale for weight mutation; the paper reports two different values.
  • mutation_rate schedule = 0.9 decaying to 0.1 over 20,000 iterations
    Exploration-to-exploitation schedule chosen by hand.
  • stop_cost = 0.035, 0.015, 0.04 per architecture
    Per-architecture early-stopping thresholds; different values make cross-architecture comparison inconsistent.
  • max_layer_dims = [12288,20,5,1], [12288,7,1], [12288,20,7,5,1]
    Manual bounds on hidden-layer sizes; the 'self-adaptive' layer mechanism only mutates within these bounds.
  • agent size = 5 architectures per agent
    Number of candidate architectures stored per agent; chosen by hand.
  • max_iter = 20,000
    Iteration budget; chosen by hand.
assumptions (5)
  • standard math Standard multilayer perceptron forward propagation with ReLU and sigmoid activations (Eq. 1).
    Used to compute costs for all candidate networks.
  • domain assumption Binary cross-entropy is the appropriate training objective (Eq. 2).
    The paper assumes minimizing cross-entropy on training data leads to good classifiers.
  • ad hoc to paper Zero-initialized weights and biases provide a usable starting point for evolution.
    All agents start with identical zero parameters; this is unusual and can break symmetry only through mutation.
  • ad hoc to paper Fitness for roulette selection is 1/(1+cost) (Eq. 4).
    Hand-chosen transformation; no justification that it preserves the true fitness ranking.
  • ad hoc to paper Pareto dominance rank counts the number of dominating solutions and preserves the correct order when sorted by cost.
    Section 2.6 admits this is not standard NSGA-II ranking.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dual-Individual Genetic Algorithm: A Dual-Individual Approach for Efficient Training of Multi-Layer Neural Networks." pith.science (2026). https://pith.science/paper/KBYQOME7

@misc{pith2026250417346,
  author       = {Pith},
  title        = {Pith review of: Dual-Individual Genetic Algorithm: A Dual-Individual Approach for Efficient Training of Multi-Layer Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KBYQOME7}},
  note         = {Machine review of arXiv:2504.17346}
}
read the original abstract

This paper introduces an enhanced Genetic Algorithm technique, which optimizes neural networks for binary image classification tasks, such as cat vs. non-cat classification. The proposed method employs only two individuals for crossover, represented by two parameter sets: Leader and Follower. The Leader focuses on exploitation, representing the primary optimal solution, while the Follower promotes exploration by preserving diversity and avoiding premature convergence. Leader and Follower are modeled as two phases or roles. The key contributions of this work are threefold: (1) a self-adaptive layer dimension mechanism that eliminates the need for manual tuning of layer architectures; (2) generates two parameter sets, leader and follower parameter sets, with 10 layer architecture configurations (5 for each set), ranked by Pareto dominance and cost post-optimization; and (3) achieved better results compared to gradient-based methods. Experimental results show that the proposed method achieves 99.04% training accuracy and 80% testing accuracy (cost = 0.06) on a three-layer network with architecture [12288, 17, 4, 1], higher performance a gradient-based approach that achieves 98% training accuracy and 80% testing accuracy (cost = 0.092) on a four-layer network with architecture [12288, 20, 7, 5, 1].

Figures

Figures reproduced from arXiv: 2504.17346 by the authors.

Figure 2
Figure 2. The framework of the proposed model ALGORITHM 1: DUAL-INDIVIDUAL GA Input: Size , X_train, Y_train, max_layer_dims, stop_cost Output: Best solution, best cost, leader_params, follower_params, leader_layers, follower_layers 1 Initialization of parameters: 2 leader_layers, leader_params = initialize_agent_network(size, X_train, Y_train, max_layer_dims) 3 follower_layers, follower_params = initialize_agent_network(size… view at source ↗
Figure 3
Figure 3. Pairing process of multi-point crossover In [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Mutation parameters over 20,000 iterations Function 4 implements a Gaussian (normal distribution) mutation. It generates random perturbations from a Gaussian distribution with mean 0, scale mutation_scale for all parameters, but a mutation_rate determines whether a parameter is actually mutated. These perturbations (mutations) are added to the agent_params. Gaussian noise mimics stochastic gradient descent (SGD) noi… view at source ↗
Figures from the paper (1 more)
Figure 7
Figure 7. Figure 7: Training cost curves for Proposed Method vs. Gradient Descent [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 18 canonical work pages

  1. [1]

    Mousavi, M

    S.S. Mousavi, M. Schukat, E. Howley, Deep Reinforcement Learning: An Overview, in: Y. Bi, S. Kapoor, R. Bhatia (Eds.), Proc. SAI Intell. Syst. Conf. IntelliSys 2016, Springer International Publishing, Cham, 2018: pp. 426–440. https://doi.org/10.1007/978-3-319-56991-8_32

    work page doi:10.1007/978-3-319-56991-8_32 2016

  2. [2]

    J. Yan, B. Luo, X. Xu, Hierarchical reinforcement learning for handling sparse rewards in multi -goal navigation, Artif. Intell. Rev. 57 (2024) 156. https://doi.org/10.1007/s10462 -024-10794-3

    doi:10.1007/s10462 2024

  3. [3]

    Rengarajan, G

    D. Rengarajan, G. Vaidya, A. Sarvesh, D. Kalathil, S. Shakkottai, Reinforcement Learning with Sparse Rewards using Guidance from Offline Demonstration, (2022). https://doi.org/10.48550/arXiv.2202.04628

    arXiv doi:10.48550/arxiv.2202.04628 2022

  4. [4]

    Andriotis, K.G

    C.P. Andriotis, K.G. Papakonstantinou, Managing engineering systems with large state and action spaces through deep reinforcement learning, Reliab. Eng. Syst. Saf. 191 (2019) 106483. https://doi.org/10.1016/j.ress.2019.04.036

    doi:10.1016/j.ress.2019.04.036 2019

  5. [5]

    J. Park, J. Lee, J. Kim, S. Han, Overcoming intermittent instability in reinforcement learning via gradient norm preservation, Inf. Sci. 709 (2025) 122081. https://doi.org/10.1016/j.ins.2025.122081

    work page arXiv 2025

  6. [6]

    Y. Song, L. Wei, Q. Yang, J. Wu, L. Xing, Y. Chen, RL -GA: A Reinforcement Learning -based Genetic Algorithm for Electromagnetic Detection Satellite Scheduling Problem, Swarm Evol. Comput. 77 (2023) 101236. https://doi.org/10.1016/j.swevo.2023.101236

    arXiv 2023

  7. [7]

    R. Chen, B. Yang, S. Li, S. Wang, A self -learning genetic algorithm based on reinforcement learning for flexible job -shop scheduling problem, Comput. Ind. Eng. 149 (2020) 106778. https://doi.org/10.1016/j.cie.2020.106778

    arXiv 2020

  8. [8]

    Sehgal, H

    A. Sehgal, H. La, S. Louis, H. Nguyen, Deep Reinforcement Learning Using Genetic Algorithm for Parameter Optimization, in: 2019 Third IEEE Int. Conf. Robot. Comput. IRC, 2019: pp. 596 –601. https://doi.org/10.1109/IRC.2019.00121

    arXiv 2019

Show all 33 references
  1. [9]

    Kazimipour, X

    B. Kazimipour, X. Li, A.K. Qin, A review of population initialization techniques for evolutionary algorithms, in: 2014 IEEE Congr. Evol. Comput. CEC, 2014: pp. 2585 –2592. https://doi.org/10.1109/CEC.2014.6900618

    2014

  2. [10]

    Michalewicz, GAs: Selected Topics, in: Z

    Z. Michalewicz, GAs: Selected Topics, in: Z. Michalewicz (Ed.), Genet. Algorithms Data Struct. Evol. Programs, Springer, Berlin, Heidelberg, 1996: pp. 57–93. https://doi.org/10.1007/978-3-662-03315-9_5

    1996 doi

  3. [11]

    Harada, E

    T. Harada, E. Alba, Parallel Genetic Algorithms: A Useful Survey, ACM Comput Surv 53 (2020) 86:1 - 86:39. https://doi.org/10.1145/3400031

    2020 doi

  4. [12]

    Z. Xiao, X. Liu, J. Xu, Q. Sun, L. Gan, Highly scalable parallel genetic algorithm on Sunway many -core processors, Future Gener. Comput. Syst. 114 (2021) 679–691. https://doi.org/10.1016/j.future.2020.08.028

    2021 doi

  5. [13]

    Harvey, The Microbial Genetic Algorithm, in: G

    I. Harvey, The Microbial Genetic Algorithm, in: G. Kampis, I. Karsai, E. Szathmáry (Eds.), Adv. Artif. Life Darwin Meets Von Neumann, Springer, Berlin, Heidelberg, 2011: pp. 126 –133. https://doi.org/10.1007/978-3-642-21314-4_16

    2011 doi

  6. [14]

    Arnold, The (1 + 1)-ES: Overvaluation, in: D.V

    D.V. Arnold, The (1 + 1)-ES: Overvaluation, in: D.V. Arnold (Ed.), Noisy Optim. Evol. Strateg., Springer US, Boston, MA, 2002: pp. 21–36. https://doi.org/10.1007/978-1-4615-1105-2_3

    2002 doi

  7. [15]

    Jägersküpper, Rigorous Runtime Analysis of the (1+1) ES: 1/5-Rule and Ellipsoidal Fitness Landscapes, in: A.H

    J. Jägersküpper, Rigorous Runtime Analysis of the (1+1) ES: 1/5-Rule and Ellipsoidal Fitness Landscapes, in: A.H. Wright, M.D. Vose, K.A. De Jong, L.M. Schmitt (Eds.), Found. Genet. Algorithms, Springer, Berlin, Heidelberg, 2005: pp. 260–281. https://doi.org/10.1007/11513575_14

    2005 doi

  8. [16]

    Arnold, D

    D.V. Arnold, D. Brauer, On the Behaviour of the (1+1) -ES for a Simple Constrained Problem, in: G. Rudolph, T. Jansen, N. Beume, S. Lucas, C. Poloni (Eds.), Parallel Probl. Solving Nat. – PPSN X, Springer, Berlin, Heidelberg, 2008: pp. 1–10. https://doi.org/10.1007/978-3-540-87700-4_1

    2008 doi

  9. [17]

    Truong, J

    T.T.N. Truong, J. Kim, Dual -Individual Genetic Algorithm: A Dual -Individual Approach for Efficient Training of Multi-Layer Neural Networks, (2025). https://doi.org/10.48550/arXiv.2504.17346

    2025 doi arXiv

  10. [18]

    https://www.coursera.org/specializations/deep - learning (accessed April 7, 2025)

    Neural Networks and Deep Learning, Coursera (n.d.). https://www.coursera.org/specializations/deep - learning (accessed April 7, 2025)

    2025

  11. [19]

    X.Z. Gao, V. Govindasamy, H. Xu, X. Wang, K. Zenger, Harmony Search Method: Theory and Applications, Comput. Intell. Neurosci. 2015 (2015) 258491. https://doi.org/10.1155/2015/258491

    2015 doi

  12. [20]

    X. Sun, H. He, L. Ma, Harmony search meta-heuristic algorithm based on the optimal sizing of wind-battery hybrid micro-grid power system with different battery technologies, J. Energy Storage 75 (2024) 109582. https://doi.org/10.1016/j.est.2023.109582

    2024

  13. [21]

    Qin, A.M

    F. Qin, A.M. Zain, K. -Q. Zhou, Harmony search algorithm and related variants: A systematic review, Swarm Evol. Comput. 74 (2022) 101126. https://doi.org/10.1016/j.swevo.2022.101126

    2022

  14. [22]

    Zamli, H.S

    K.Z. Zamli, H.S. Alhadawi, F. Din, Utilizing the roulette wheel based social network search algorithm for substitution box construction and optimization, Neural Comput. Appl. 35 (2023) 4051 –4071. https://doi.org/10.1007/s00521-022-07899-7

    2023 doi

  15. [23]

    Pham, N.T

    V.H.S. Pham, N.T. Nguyen Dang, V.N. Nguyen, Enhancing engineering optimization using hybrid sine cosine algorithm with Roulette wheel selection and opposition -based learning, Sci. Rep. 14 (2024) 694. https://doi.org/10.1038/s41598-024-51343-w

    2024 doi

  16. [24]

    An Added Level of Sophistication, in: Pract. Genet. Algorithms, John Wiley & Sons, Ltd, 2003: pp. 95 –

    2003

  17. [25]

    Bejarano, H.E

    L.A. Bejarano, H.E. Espitia, C.E. Montenegro, Clustering Analysis for the Pareto Optimal Front in Multi - Objective Optimization, Computation 10 (2022) 37. https://doi.org/10.3390/computation10030037

    2022 doi

  18. [26]

    K. Deb, A. Pratap, S. Agarwal, T. Meyarivan, A fast and elitist multiobjective genetic algorithm: NSGA - II, IEEE Trans. Evol. Comput. 6 (2002) 182–197. https://doi.org/10.1109/4235.996017

    2002

  19. [27]

    H. Sun, L. Shen, Q. Zhong, L. Ding, S. Chen, J. Sun, J. Li, G. Sun, D. Tao, AdaSAM: Boosting sharpness- aware minimization with adaptive learning rate and momentum for training deep neural networks, Neural Netw. 169 (2024) 506–519. https://doi.org/10.1016/j.neunet.2023.10.044

    2024 doi

  20. [28]

    Y. Tian, Y. Zhang, H. Zhang, Recent Advances in Stochastic Gradient Descent in Deep Learning, Mathematics 11 (2023) 682. https://doi.org/10.3390/math11030682

    2023 doi

  21. [29]

    Kunang, S

    Y.N. Kunang, S. Nurmaini, D. Stiawan, B.Y. Suprapto, Attack classification of an intrusion detection system using deep learning and hyperparameter optimization, J. Inf. Secur. Appl. 58 (2021) 102804. https://doi.org/10.1016/j.jisa.2021.102804

    2021

  22. [30]

    Montesinos López, A

    O.A. Montesinos López, A. Montesinos López, J. Crossa, Overfitting, Model Tuning, and Evaluation of Prediction Performance, in: O.A. Montesinos López, A. Montesinos López, J. Crossa (Eds.), Multivar. Stat. Mach. Learn. Methods Genomic Predict., Spring er International Publishi...

    2022 doi

  23. [31]

    Bejani, M

    M.M. Bejani, M. Ghatee, A systematic review on overfitting control in shallow and deep neural networks, Artif. Intell. Rev. 54 (2021) 6391–6438. https://doi.org/10.1007/s10462-021-09975-1

    2021 doi

  24. [32]

    Y. Tian, Y. Zhang, A comprehensive survey on regularization strategies in machine learning, Inf. Fusion 80 (2022) 146–166. https://doi.org/10.1016/j.inffus.2021.11.005

    2022 doi

  25. [150]

    https://doi.org/10.1002/0471671746.ch5

    doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.