Pith. sign in

REVIEW 3 major objections 5 minor 26 references

Meta-Black-Box-Optimization through Offline Q-function Learning

T0 review · 3 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 The paper claims that dynamic configuration of a black-box optimizer can be meta-learned entirely offline, matching or beating online reinforcement-learning baselines on unseen problems while using less than half the training budget.

desk verdict First offline MetaBBO paper with honest experiments and a sensible decomposition-plus-CQL design; the headline efficiency claim, though, only works if you ignore the cost of pre-training the behavior policies that generate its own dataset. read the letter →

arxiv 2505.02010 v1 pith:3WNPMRHO submitted 2025-05-04 cs.NE cs.LG

Zeyuan Ma , Zhiguang Cao , Zhou Jiang , Hongshu Guo , Yue-Jiao Gong This is my paper · ORCID
classification cs.NEcs.LG
keywords metablack-boxoptimizationofflinereinforcementlearningdynamicalgorithmconfigurationQ-functiondecompositionconservativeQ-learningMambastatespacemodelBBOBbenchmarkneuroevolution
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

Meta-black-box-optimization normally learns its controller by interacting online with the optimizer-environment, which is slow because a single run takes hundreds of optimization steps. This paper claims that the same quality of control can be learned offline: given a fixed dataset of past configuration trajectories, an agent can match or slightly beat online-trained agents on unseen benchmark problems while consuming at most half the training budget. The proposed system, Q-Mamba, rewrites the configuration choice as a long sequence of per-parameter decisions, learns a decomposed Q-function for each parameter with a conservative regularization term that keeps out-of-distribution actions from being overvalued, and processes the whole sequence with a Mamba state-space model. The practical bet is that expensive online data collection in algorithm configuration can be replaced by a single offline dataset that mixes expert and random trajectories, with no loss in downstream optimization quality.

What carries the argument

The load-bearing object is the decomposed, discretized Q-function: each of the K hyper-parameters of the low-level optimizer is mapped to M=16 action bins (5-bit tokens), and the Q-learner predicts the value of bin j for parameter i autoregressively, conditioned on the optimization state and the previously selected bins, Q(a_i | s, a_{1:i}); the per-dimension Bellman backup ties dimension i to the max over dimension i+1, and the last dimension closes the loop with the reward and next state. Around this, three mechanisms carry the argument: the E&E dataset, 10,000 trajectories split evenly between rollouts of three pre-trained MetaBBO policies and a random configuration policy, which supplies both exploitation quality and exploration coverage; the compositional Q-loss, whose three branches are the TD error for the selected bin, a weighted (beta=10) TD error on the last dimension, and a CQL-style conservative term (lambda=1) pushing unselected bins toward zero; and the Mamba block, a selective state-space model whose input-dependent parameters and hardware-aware parallel scan make long-sequence Q-learning efficient. The appendix's proof that the decomposed Bellman updates equal the full-action Bellman update is what legitimizes treating configuration as a sequence problem.

What would settle it

Two checks would settle the claim. Directly: count the distinct discretized configuration bins (out of 16^K for K=16) that appear in the exploitation half of the dataset, then compare how often the deployed policy selects bins absent from the dataset; if unvisited bins are chosen frequently and systematically underperform visited alternatives in brute-force evaluation, the coverage premise fails. Experimentally: train Q-Mamba on trajectories restricted to one sub-region of a continuous hyper-parameter (for example DE scale factor F <= 0.5), then test on the complementary region; if its argmax bin there does not match the best bin found by exhaustive evaluation, the offline Q-function has not learned the missing region.

Watch

Extended reading notes

Core claim

The paper's central claim is that dynamic algorithm configuration does not need online interaction: a meta-level policy trained purely by offline Q-learning on a pre-collected dataset matches or exceeds online RL baselines (RLPSO, LDE, GLEET) on held-out BBOB instances, and does so at roughly half the training time (about 13 hours versus 25-28 hours). The discovery that makes this work is that the joint configuration action should be decomposed into a sequence of per-hyperparameter Q-functions, one prediction step per parameter in a fixed order, so that learning the value of the massive joint action space is replaced by learning K small autoregressive classification problems over 16 discretized bins each. Training uses a compositional loss that combines the decomposed Bellman backup with a conservative Q-learning term that drives Q-values of unselected bins toward zero, and the data comes from an exploration-exploitation balanced dataset in which half the trajectories are rollouts of pre-trained MetaBBO policies and half are random configurations. A proof in the appendix shows the decomposed backups are consistent with the full Bellman backup, so optimizing the per-dimension Q-functions optimizes the original MDP. The same policy transfers zero-shot to neuroevolution on Mujoco control tasks, where it performs comparably to the online baselines despite having been trained only on synthetic benchmark functions.

Load-bearing premise

The offline dataset must actually cover enough of the configuration space, the regions expert policies seldom visit as well as the ones they favor, for the learned Q-values to rank actions correctly on out-of-distribution tasks like neuroevolution; if the behavior policies under-cover the space, the offline Q-function cannot recover good decisions there.

Editorial extensions

If this is right

  • If the results hold, offline data alone yields a meta-level configuration policy whose mean accumulated performance improvement on held-out BBOB instances is competitive with or higher than every online and offline baseline across all three optimizer complexities (K=3, 10, 16).
  • Meta-training time drops to roughly half that of the online baselines (about 13 hours versus 25-28 hours), with no extra inference cost, making learned configuration practical when each optimization run is itself expensive.
  • Zero-shot transfer works: a controller trained only on synthetic BBOB functions of at most 50 dimensions can configure the evolution of MLP policies with thousands of parameters on Mujoco continuous-control tasks at a level comparable to the online baselines.
  • The two loss coefficients carry weight: removing the conservative regularization (lambda=0) or unweighting the last action dimension (beta=1) both measurably reduce performance, and a 50/50 mix of expert and random trajectories beats either pure exploitation or pure exploration data.
  • Control granularity is not the bottleneck: increasing action bins from 16 up to 512 does not improve performance, consistent with the tested evolutionary operators being insensitive to small parameter changes.

Reading between the lines

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

  • The efficiency gain is amortized, not free: the exploitation half of the dataset is produced by pre-training three online MetaBBO policies, so an end-to-end cost accounting that includes that upstream training is the paper's implicit next step; the paper also notes Q-Mamba must be retrained from scratch for a new optimizer because no algorithm-identity features are provided to the Q-learner.
  • The decomposition recipe is generic beyond evolutionary computation: any control problem with a factored action space and a long decision horizon (scheduling, circuit or prompt configuration) could use the same offline Q-decomposition and Mamba backbone, but the paper only demonstrates it on MetaBBO and explicitly leaves validation on other RL domains to future work.
  • The optimal data mix likely tracks behavior-policy quality: because mu=0.5 beats both extremes, a dataset-quality curve varying mu against the strength of the exploitation policies would reveal whether the 50/50 choice is intrinsic to dynamic algorithm configuration or an artifact of these three baselines.
  • Discretization caps the achievable control: the 16-bin uniform grid makes parameter values inside a bin indistinguishable to the policy, so for hyper-parameters with sharp performance cliffs a post-hoc refinement of the selected bin's interior would be a natural extension the paper does not test.
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

3 major / 5 minor

Summary. The paper proposes Q-Mamba, an offline meta-black-box-optimization (MetaBBO) framework for dynamic algorithm configuration (DAC). The method transforms DAC into a long-sequence decision process, decomposes the Q-function over the low-level algorithm's hyperparameters, and learns the decomposed Q-functions from a pre-collected dataset of DAC trajectories using a Mamba-based sequence model with a conservative Q-learning loss. The authors compare Q-Mamba with online MetaBBO baselines (RLPSO, LDE, GLEET) and offline RL baselines (DT, DeMa, QDT, QT, Q-Transformer) on BBOB training/test instances, and also evaluate zero-shot transfer to neuroevolution tasks. They report competitive or superior optimization performance and claim that Q-Mamba consumes at most half the training budget of the online baselines.

Significance. If the efficiency claim is corrected to account for all pipeline costs, the paper would still be a valuable first step in offline MetaBBO: it introduces a plausible combination of Q-function decomposition, conservative regularization, and a Mamba sequence model, and it provides code, ablations, and transfer experiments. The in-distribution results show that an offline policy can roughly match the performance of online baselines, which is a useful existence proof. However, the headline efficiency contribution is currently undercut by an incomplete cost accounting and by an arithmetic inconsistency even within the paper's own Table 1, so the significance as stated is not yet established.

major comments (3)
  1. [§4.2, Table 1, §5.2(i), Abstract] The claim that Q-Mamba consumes at most half the training budget of online baselines omits the cost of building the E&E dataset. Section 4.2 states that the dataset is collected by pre-training RLPSO, LDE, and GLEET on the target algorithm/problem distribution and then rolling them out; these pre-training runs are the same online runs whose times are listed in Table 1 (28h, 28h, 25h). A full-cost accounting therefore gives roughly 28+28+25+13=94h for the Q-Mamba pipeline, which is over three times the cost of a single online baseline. Even if only one behavior policy were used, 28h+13h=41h exceeds the half-budget bound of 12.5h for GLEET. Moreover, the arithmetic fails even with Q-Mamba's own reported 13h, since 13h is not at most half of GLEET's 25h. The efficiency comparison should either include behavior-policy training time, with the amortization scenario explicitly stated, or the claims in the abstract and Section 5.2 should be restricted to 'policy-optimizer training time' rather than total training budget.
  2. [Table 1, §5.2] The results support 'competitive' but not 'superior' performance, because the reported differences between Q-Mamba and the best online baselines are within the reported error bars on all three low-level algorithms. For example, on Alg1 the mean performance of Q-Mamba is 9.973E-01 ± 2.441E-03 while RLPSO is 9.953E-01 ± 3.322E-03, and on Alg2 the interval for Q-Mamba (9.950E-01 ± 9.981E-03) fully overlaps that of RLPSO (9.914E-01 ± 4.497E-03). No significance tests or effect sizes are reported. The paper should add a paired or per-instance significance test across the 19 runs, or soften the superiority claim to a competitiveness claim.
  3. [§4.3, Eq. (5), Appendix A] Equation (5) uses the same network parameters for the prediction and the target in the decomposed Bellman backups. The targets max_j Q^t_{i+1,j} and r_t + γ max_j Q^{t+1}_{1,j} are outputs of the same Q-function being trained, and the paper does not state whether stop-gradient or a target network is applied. Without stop-gradient, the loss for i<K is not a standard Bellman backup for Q^t_i but a consistency constraint that could in principle be minimized by driving Q^t_i and max_j Q^t_{i+1,j} to a common constant, with the reward signal entering only through the final action dimension. The convergence argument in Appendix A assumes an exact optimal Q-function update, not the gradient dynamics of Eq. (5). Please state explicitly whether gradients are blocked on the target terms, and if they are not, explain why the network does not exhibit the collapse suggested by the loss structure.
minor comments (5)
  1. [§5.2(i)] There is a typo: 'his is especially appealing' should read 'This is especially appealing'.
  2. [Table 4, rows s7–s8] The description of s7 contains 'potion' instead of 'portion', and the definition of s8 uses s_t both as a stagnation counter and as the state symbol, which is confusing; please rename one of the two quantities.
  3. [§4.2] The random strategy used to collect the exploration trajectories is not specified beyond 'randomly control the hyper-parameters'; please state whether hyperparameter values are drawn uniformly over their ranges, over the discretized bins, or in some other way, since this affects the coverage of the offline dataset.
  4. [Table 1, last row] The timing comparison would be more meaningful if the hardware and software environment (GPU/CPU, batch size, number of workers) were reported, and if the same timing protocol were used for all baselines (including any evaluation or checkpointing overhead).
  5. [§4.4] The architecture details for the Mamba-based Q-learner are minimal ('we use the mamba-block in Mamba repo, with default settings'); for reproducibility, please report the number of layers, hidden dimension, state dimension, and total parameter count.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: Q-Mamba's effectiveness is tested on held-out BBOB instances and neuroevolution tasks, and its conservative Q-learning objective can improve over the behavior policies that generated the dataset; the mild self-citation (GLEET as data generator and baseline) is not load-bearing, and the efficiency claim's excluded data-collection cost is a cost-accounting caveat…

full rationale

The central effectiveness result does not reduce to the paper's inputs by construction. The E&E dataset (Section 4.2) mixes trajectories from RLPSO, LDE, GLEET, and a random policy on 16 BBOB training instances, and Q-Mamba is then tested on the 8 held-out BBOB instances (Section 5.2, Table 1) and on out-of-distribution neuroevolution tasks (Section 5.3). Because the test problems are unseen and the training objective in Eq. (5) is a conservative Q-learning loss rather than behavior cloning, Q-Mamba is not constrained to reproduce the generating policies; it can and does improve on them. The Q-function decomposition (Eq. (2), Appendix A) is a proof of equivalence to the standard Bellman backup and is explicitly credited to prior work (SDQN, Q-Transformer), so it is not a self-citational or definitional shortcut. The use of the authors' own GLEET as both a data-collection policy and a comparison baseline is a minor self-citation, but it is not load-bearing because the held-out evaluation and the inclusion of RLPSO/LDE/random trajectories keep the comparison informative. The efficiency claim that Q-Mamba consumes 'at most half training budget' (Section 5.2) is subject to a legitimate full-cost-accounting objection: Table 1's 13h does not include the pre-training time of the online MetaBBO methods that Section 4.2 requires to build the E&E dataset; however, this is a comparison/accounting caveat, not a circularity of the derivation chain. The paper itself acknowledges a concrete limitation in the Conclusion: 'Q-Mamba is trained for a given BBO algorithm and requires re-training for other algorithms.' Overall, no prediction or first-principles result in the paper is equivalent to its inputs by construction.

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

The central claim rests on a handful of modeling and dataset assumptions, all typical for offline RL applied to a new domain. The free parameters beta, lambda, mu, and M are explicitly ablated, which is good practice, but gamma is missing. No new physical or ontological entities are introduced; Q-Mamba is a software architecture.

free parameters (5)
  • beta (beta) = 10
    Weight on the last-action-dimension TD error in the compositional Q-loss (Eq. 5); chosen by hand to reinforce the final dimension, ablated in Table 2.
  • lambda (lambda) = 1
    Weight on the conservative regularization that pushes unselected action bins toward 0 (Eq. 5); chosen by hand, ablated in Table 2.
  • mu (mu) = 0.5
    Fraction of exploitation trajectories in the E&E dataset (Section 4.2); chosen by hand, ablated in Table 3.
  • M (action bins per hyperparameter) = 16
    Number of discretization bins for each hyperparameter (Section 4.4); shown to be robust versus 32 to 512 bins on Alg0 in Appendix E.1, but assumed adequate for Alg1 and Alg2.
  • discount factor gamma = not reported in text
    Appears in Eq. (1) and Eq. (5) but its value is never stated in the paper, a reproducibility gap.
assumptions (5)
  • standard math The decomposed Q-function optimization (Eq. 2) is equivalent to full-action Q-function optimization.
    The paper provides a brief proof in Appendix A that rests on the inverse maximization operation; this is a known result from Q-Transformer (Chebotar et al., 2023) and is not machine-checked.
  • domain assumption The 9-dimensional handcrafted state st (Appendix B, Table 4) is a sufficient statistic for dynamic algorithm configuration decisions.
    The state is inherited from GLEET (Ma et al., 2024b) and is not validated or ablated in this paper; if it omits relevant optimization information, the learned Q-function cannot compensate.
  • domain assumption The E&E dataset collected from three pre-trained online baselines and a random policy provides sufficient coverage of the DAC state-action space.
    Section 4.2 assumes that mixing 50% baseline trajectories with 50% random trajectories yields a dataset rich enough for offline learning; the ablations vary only the mixing ratio, not the choice of behavior policies.
  • domain assumption Uniform discretization into M=16 bins preserves near-optimal control for continuous hyperparameters.
    Appendix E.1 shows 16 bins is comparable to up to 512 bins on Alg0, and the paper assumes this transfers to Alg1 and Alg2 without dedicated experiments.
  • domain assumption The surrogate optimum used to normalize rewards is close to the true optimum on training problems.
    Section 5.1 states f* is replaced by a surrogate obtained by running an advanced BBO algorithm; if the surrogate is poor, the reward normalization biases the Perf metric.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Meta-Black-Box-Optimization through Offline Q-function Learning." pith.science (2026). https://pith.science/paper/3WNPMRHO

@misc{pith2026250502010,
  author       = {Pith},
  title        = {Pith review of: Meta-Black-Box-Optimization through Offline Q-function Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3WNPMRHO}},
  note         = {Machine review of arXiv:2505.02010}
}
read the original abstract

Recent progress in Meta-Black-Box-Optimization (MetaBBO) has demonstrated that using RL to learn a meta-level policy for dynamic algorithm configuration (DAC) over an optimization task distribution could significantly enhance the performance of the low-level BBO algorithm. However, the online learning paradigms in existing works makes the efficiency of MetaBBO problematic. To address this, we propose an offline learning-based MetaBBO framework in this paper, termed Q-Mamba, to attain both effectiveness and efficiency in MetaBBO. Specifically, we first transform DAC task into long-sequence decision process. This allows us further introduce an effective Q-function decomposition mechanism to reduce the learning difficulty within the intricate algorithm configuration space. Under this setting, we propose three novel designs to meta-learn DAC policy from offline data: we first propose a novel collection strategy for constructing offline DAC experiences dataset with balanced exploration and exploitation. We then establish a decomposition-based Q-loss that incorporates conservative Q-learning to promote stable offline learning from the offline dataset. To further improve the offline learning efficiency, we equip our work with a Mamba architecture which helps long-sequence learning effectiveness and efficiency by selective state model and hardware-aware parallel scan respectively. Through extensive benchmarking, we observe that Q-Mamba achieves competitive or even superior performance to prior online/offline baselines, while significantly improving the training efficiency of existing online baselines. We provide sourcecodes of Q-Mamba at https://github.com/MetaEvo/Q-Mamba.

Figures

Figures reproduced from arXiv: 2505.02010 by the authors.

Figure 1
Figure 1. The workflow of the Mamba-based Q-Learner. The forward process of the neural network is similar with the Recurrent Neural Network. At each time step, the Q-function of each decomposed action dimension is outputted by conditioning the current state and selected action bins of the previous action dimensions. The environment transition is executed once all action dimensions are outputted. pectation training objective E… view at source ↗
Figure 2
Figure 2. Zero shot performance of Q-Mamba and online MetaBBO baselines on neuroevolution tasks. a population of neural networks according to their perfor￾mance on a specific machine learning task, i.e., classifica￾tion, robotic control (Galvan & Mooney ´ , 2021). Specifically, we consider continuous control tasks in Mujoco (Todorov et al., 2012). We optimize a 2-layer MLP policy for each task by Q-Mamba and other baselines t… view at source ↗
Figure 3
Figure 3. Fitness landscapes of functions in BBOB train set when dimension is set to 2. 4 2 0 2 4 4 2 0 2 4 1 2 3 4 1e7 (a) f2 4 2 0 2 4 4 2 0 2 4 500 1000 1500 2000 2500 3000 (b) f3 4 2 0 2 4 4 2 0 2 4 0 20 40 60 80 100 (c) f5 4 2 0 2 4 4 2 0 2 4 50 100 150 200 (d) f6 4 2 0 2 4 4 2 0 2 4 500 1000 1500 2000 2500 3000 3500 (e) f7 4 2 0 2 4 4 2 0 2 4 20000 40000 60000 80000 100000 120000 140000 160000 (f) f8 4 2 0 2 4 4 2 0 2 4… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Fitness landscapes of functions in BBOB test set when dimension is set to 2. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: The performances of Q-Mamba trained with different action bin granularities. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 13 canonical work pages

  1. [2]

    4 2 0 2 4 4 2 0 2 4 1 2 3 4 1e7 (a)f2 4 2 0 2 4 4 2 0 2 4 500 1000 1500 2000 2500 3000 (b)f3 4 2 0 2 4 4 2 0 2 4 0 20 40 60 80 100 (c)f5 4 2 0 2 4 4 2 0 2 4 50 100 150 200 (d)f6 4 2 0 2 4 4 2 0 2 4 500 1000 1500 2000 2500 3000 3500 (e)f7 4 2 0 2 4 4 2 0 2 4 20000 40000 60000 80000 100000 120000 140000 160000 (f)f8 4 2 0 2 4 4 2 0 2 4 20000 40000 60000 800...

    work page 2000

  2. [4]

    States Notes Problem Property st 1 mean xi,xj∈Xt ||xi−xj||2 Average distance between any pair of individuals in current population

    Formulations of state features. States Notes Problem Property st 1 mean xi,xj∈Xt ||xi−xj||2 Average distance between any pair of individuals in current population. st 2 mean xi∈Xt ||xi−x∗,t||2 Average distance between each individual and the best individual in t-th generation. st 3 mean xi∈Xt ||xi−x∗||2 Average distance between each individual and the bes...

    work page 1997

  3. [5]

    Overview of the BBOB testsuites. Problem Functions Dimensions Separable functions f1 Sphere Function 50 f2 Ellipsoidal Function 5 f3 Rastrigin Function 5 f4 Buche-Rastrigin Function 10 f5 Linear Slope 50 Functions with low or moderate conditioning f6 Attractive Sector Function 5 f7 Step Ellipsoidal Function 20 f8 Rosenbrock Function, original 10 f9 Rosenb...

    work page 2000

  4. [7]

    Large language model agent for hyper-parameter optimization

    Liu, S., Gao, C., and Li, Y . Large language model agent for hyper-parameter optimization. arXiv preprint arXiv:2402.01881,

    arXiv

  5. [9]

    Playing atari with deep reinforcement learning

    Mnih, V . Playing atari with deep reinforcement learning. arXiv preprint arXiv:1312.5602,

    arXiv

  6. [12]

    Proximal policy optimization algorithms

    Schulman, J., Wolski, F., Dhariwal, P., Radford, A., and Klimov, O. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347,

    arXiv

  7. [13]

    Reinforced in-context black-box opti- mization

    Song, L., Gao, C., Xue, K., Wu, C., Li, D., Hao, J., Zhang, Z., and Qian, C. Reinforced in-context black-box opti- mization. arXiv preprint arXiv:2402.17423,

    arXiv

  8. [15]

    and Fukunaga, A

    Tanabe, R. and Fukunaga, A. S. Improving the search per- formance of shade using linear population size reduction. In 2014 IEEE Congress on Evolutionary Computation ,

    work page 2014

Show all 26 references
  1. [17]

    Decn: Evolution inspired deep convolution network for black-box optimization

    Wu, K., Li, X., Liu, P., and Liu, J. Decn: Evolution inspired deep convolution network for black-box optimization. arXiv preprint arXiv:2304.09599,

    arXiv

  2. [18]

    Automated metaheuristic algorithm design with au- toregressive learning

    Zhao, Q., Liu, T., Yan, B., Duan, Q., Yang, J., and Shi, Y . Automated metaheuristic algorithm design with au- toregressive learning. arXiv preprint arXiv:2405.03419,

    arXiv

  3. [20]

    with Linear Population Size Reduction (LPSR) (Tanabe & Fukunaga, 2014). The mutation operator DE/current-to-rand/1 is formulated as: x′ i =xi +F 1(xr1−xi) +F 2(xr2−xr3) (10) wherexr· are randomly chosen solutions andF 1,F 2∈ [0, 1] are two controllable parameters. The Exponent...

    2014

  4. [23]

    (17) whereηc∈{ 1, 2, 3} is controllable parameter andu∈ [0, 1] is random number

    and Tournament selection (Goldberg & Deb, 1991): x′ i = 0.5· [(1∓β)xi + (1±β)xr1], whereβ = ( (2u) 1 1+ηc− 1, ifu≤ 0.5; ( 1 2−2u ) 1 1+ηc, ifu> 0.5. (17) whereηc∈{ 1, 2, 3} is controllable parameter andu∈ [0, 1] is random number. Similar to MPX, SBX also uses an action Xrsbx t...

    1991

  5. [24]

    The Exponential crossover formulated as Eq

    where the DE/rand/2 mutation operator is: x′ i =xr1 +F 13(xr2−xr3) +F 23(xr4−xr5) (18) wherexr· are randomly selected solutions andF 13,F 23∈ [0, 1] are controllable parameters for the third sub-population. The Exponential crossover formulated as Eq. (11) is used in this sub-p...

    1997

  6. [1991]

    and Dao, T

    Gu, A. and Dao, T. Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752,

    arXiv

  7. [1992]

    clip”, “rand

    accompanying with the Roulette selection (Holland, 1992). MPX crossover is formulated as: x′ i = ( x′ r1,j, ifrandj <Cr 1 x′ i,j, otherwise ,j = 1,··· ,Dim (12) whererandj∈ [0, 1] are random numbers, Cr 1 is a controllable parameter and xr1 is a random solution. The sample met...

    1992

  8. [1995]

    C., Pearson, D

    Dobnikar, A., Steele, N. C., Pearson, D. W., Albrecht, R. F., Deb, K., and Agrawal, S. A niched-penalty approach for constraint handling in genetic algorithms. In Artificial Neural Nets and Genetic Algorithms: Proceedings of the International Conference in Portoroˇz, Slovenia, 1999,

    1999

  9. [1997]

    P., Madhavan, V ., Conti, E., Lehman, J., Stanley, K

    Such, F. P., Madhavan, V ., Conti, E., Lehman, J., Stanley, K. O., and Clune, J. Deep neuroevolution: Genetic algo- rithms are a competitive alternative for training deep neu- ral networks for reinforcement learning. arXiv preprint arXiv:1712.06567,

    arXiv

  10. [1999]

    The Polynomial mutation is as follow: x′′ i = ( x′ i + ((2u) 1 1+ηm− 1)(x′ i−lb), ifu≤ 0.5; x′ i + (1− (2− 2u) 1 1+ηm )(ub−x′ i), ifu> 0.5

    accompanying with the Roulette selection (Holland, 1992). The Polynomial mutation is as follow: x′′ i = ( x′ i + ((2u) 1 1+ηm− 1)(x′ i−lb), ifu≤ 0.5; x′ i + (1− (2− 2u) 1 1+ηm )(ub−x′ i), ifu> 0.5. (16) whereηm∈{ 1, 2, 3} is a controllable parameter,u∈ [0, 1] is a random numbe...

    1992

  11. [2013]

    W., Hadi, A

    Mohamed, A. W., Hadi, A. A., Mohamed, A. K., Agrawal, P., Kumar, A., and Suganthan, P. N. Problem definitions and evaluation criteria for the cec 2021 on single objective bound constrained numerical optimization. In Proceed- ings of the IEEE Congress of Evolutionary Computation,

    2021

  12. [2014]

    Mujoco: A physics engine for model-based control

    Todorov, E., Erez, T., and Tassa, Y . Mujoco: A physics engine for model-based control. In 2012 IEEE/RSJ in- ternational conference on intelligent robots and systems,

    2012

  13. [2017]

    Is mamba compatible with trajectory optimization in offline reinforcement learning? arXiv preprint arXiv:2405.12094, 2024a

    Dai, Y ., Ma, O., Zhang, L., Liang, X., Hu, S., Wang, M., Ji, S., Huang, J., and Shen, L. Is mamba compatible with trajectory optimization in offline reinforcement learning? arXiv preprint arXiv:2405.12094, 2024a. Dai, Y ., Ma, O., Zhang, L., Liang, X., Hu, S., Wang, M., Ji, S...

    arXiv

  14. [2020]

    Discovering attention-based genetic algorithms via meta-black-box optimization

    Lange, R., Schaul, T., Chen, Y ., Lu, C., Zahavy, T., Dal- ibard, V ., and Flennerhag, S. Discovering attention-based genetic algorithms via meta-black-box optimization. In Proceedings of the Genetic and Evolutionary Computa- tion Conference, 2023a. Lange, R. T., Schaul, T., C...

    2005 arXiv

  15. [2021]

    Decision mamba: Reinforcement learning via sequence modeling with selective state spaces

    Ota, T. Decision mamba: Reinforcement learning via sequence modeling with selective state spaces. arXiv preprint arXiv:2403.19925,

    arXiv

  16. [2022]

    Deep reinforcement learning for dynamic algorithm selection: A proof-of-principle study on differential evolution

    Guo, H., Ma, Y ., Ma, Z., Chen, J., Zhang, X., Cao, Z., Zhang, J., and Gong, Y .-J. Deep reinforcement learning for dynamic algorithm selection: A proof-of-principle study on differential evolution. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2024a. Guo, H., M...

    arXiv

  17. [2023]

    Neu- ral exploratory landscape analysis

    Ma, Z., Chen, J., Guo, H., and Gong, Y .-J. Neu- ral exploratory landscape analysis. arXiv preprint arXiv:2408.10672, 2024a. Ma, Z., Chen, J., Guo, H., Ma, Y ., and Gong, Y .-J. Auto- configuring exploration-exploitation tradeoff in evolution- ary computation via deep reinforc...

    arXiv

  18. [2024]

    Q-value regularized transformer for offline reinforcement learning

    Hu, S., Fan, Z., Huang, C., Shen, L., Zhang, Y ., Wang, Y ., and Tao, D. Q-value regularized transformer for offline reinforcement learning. arXiv preprint arXiv:2405.17098,

    arXiv

Pith tools

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