Pith. sign in

REVIEW 3 major objections 5 minor 55 references

Discovering Quality-Diversity Algorithms via Meta-Black-Box Optimization

T0 review · 3 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read Meta-learning can automatically discover new Quality-Diversity algorithms whose learned local competition rules match or beat hand-designed baselines and transfer to novel domains like robot control.

desk verdict Solid empirical meta-learning paper whose SOTA claim about Quality-Diversity is undercut by its own QD-objective results; the emergent-diversity finding is the real contribution. read the letter →

arxiv 2502.02190 v1 pith:6CONZLRU submitted 2025-02-04 cs.NE cs.LG

Maxence Faldor , Robert Tjarko Lange , Antoine Cully This is my paper · ORCID
classification cs.NEcs.LG
keywords quality-diversitymeta-black-boxoptimizationlearnedalgorithmstransformerlocalcompetitiongeneticrobotcontrolBBOBbenchmarks
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

Quality-Diversity algorithms such as MAP-Elites and Dominated Novelty Search work by making individuals compete locally in a descriptor space, but their competition rules are hand-designed heuristics. This paper asks whether those rules can themselves be discovered by meta-learning: it parameterizes the competition function as a small transformer and evolves its weights with a meta-evolutionary strategy on a family of black-box optimization functions. The resulting Learned Quality-Diversity algorithms match or outperform established baselines on the training tasks, scale to larger populations and higher dimensions than seen in training, and transfer to robot control domains with hand-crafted behavioral descriptors. A fitness-only variant spontaneously maintains a diverse population and beats a standard genetic algorithm, which the authors read as meta-learning rediscovering diversity as instrumental for peak performance.

What carries the argument

The central object is the Learned Quality-Diversity (LQD) competition function: a transformer that maps a population's standardized fitness and descriptor features to scalar competition scores used for truncation selection. The transformer's dot-product self-attention is permutation-equivariant, so the competition outcome does not depend on population ordering; the weights are optimized by an outer Separable CMA-ES meta-evolutionary scheme over a distribution of noisy, rotated BBOB tasks whose descriptors come from random linear projections of genotypes (justified by the Johnson-Lindenstrauss lemma). This machinery turns 'design a QD algorithm' into 'search over competition-function weights'.

What would settle it

Take the meta-trained LQD (F), evaluate it on a new robot task with a categorical behavior descriptor such as gait type rather than a continuous vector, and compare with the paper's random-descriptor ablation; if performance drops to genetic-algorithm level whenever the descriptor space is not a linear image of genotype space, the transfer claim is refuted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a competition rule for Quality-Diversity optimization does not have to be invented by hand: it can be found by meta-optimizing a permutation-equivariant neural network that reads fitness and descriptor pairs and outputs a competition score for each individual. When this learned rule is inserted into the standard genetic-algorithm loop in place of global competition, it produces algorithms that outperform MAP-Elites on almost all tested configurations, match or exceed the best novelty-seeking baseline on out-of-distribution functions, and, on six aggregated robot control tasks, significantly outperform every baseline except the genetic algorithm. The same learned rule, trained only to maximize fitness, keeps the population far more diverse than a standard GA, so the paper concludes that maintaining diversity is rediscovered as an instrumental objective.

Load-bearing premise

The paper assumes that a competition rule learned in a descriptor space built from random linear projections of genotypes will still work when the descriptor space is a hand-coded behavior space like foot-contact patterns.

Editorial extensions

If this is right

  • Meta-learning can replace grid-based and nearest-neighbor competition heuristics with learned rules without sacrificing performance.
  • Learned competition rules trained on 2- to 12-dimensional BBOB problems continue to work on populations up to 1024 and dimensions up to 32, suggesting the rules capture principles rather than memorized task details.
  • The learned rule transfers to robot control with hand-crafted descriptors such as foot-contact patterns and velocities, beating all baselines except the genetic algorithm in aggregate.
  • A fitness-only LQD maintains diversity as a byproduct, implying that local competition alone can prevent premature convergence.
  • Jointly optimizing fitness and novelty remains harder: LQD (F+N) approaches but does not consistently beat Dominated Novelty Search.

Reading between the lines

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

  • Editorial extension: if the transfer result is robust, meta-discovered competition rules could replace hand-tuned QD heuristics in any domain that already has a meaningful descriptor space, since training never needs that domain's descriptors.
  • Editorial extension: the random-projection training distribution suggests a concrete stress test: vary the dimension or distribution of the projection during meta-training and measure whether robot-control transfer improves; the Johnson-Lindenstrauss justification predicts it should be insensitive above a threshold.
  • Editorial extension: the fact that LQD with random descriptors collapses to GA behavior implies the learned rule is not a general evolutionary optimizer but a descriptor-guided one; a stronger claim would require the rule to exploit genotype geometry alone, which the paper does not assert.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces Learned Quality-Diversity (LQD), a framework in which the competition step of a population-based optimizer is parameterized by a transformer and trained with meta-black-box optimization on a distribution of BBOB functions. The authors train three variants optimizing fitness, novelty, and a QD score, and evaluate them on held-out BBOB functions, on scaling tests, and on robot control tasks. The main claims are that LQD discovers sophisticated local competition rules, that it generalizes well beyond its training distribution, and that it matches or advances the state of the art in Quality-Diversity optimization.

Significance. The paper is methodologically ambitious: treating the competition rule itself as a learnable, permutation-equivariant function and meta-optimizing it with Sep-CMA-ES is a genuinely new step for the QD field. The empirical protocol is solid in several respects: 32 random seeds, Wilcoxon–Mann–Whitney tests with Holm–Bonferroni correction, held-out BBOB functions, and out-of-distribution robot control tasks. The descriptor-ablation in Section 6.3 is a useful control showing that LQD uses descriptor information rather than merely exploiting extra input dimensions. The emergent-diversity observation (Section 6.1) is interesting and well supported by the reported novelty scores. If the central QD advance were established, this would be a significant contribution; however, as detailed below, the paper's own results undercut the strongest framing.

major comments (3)
  1. [Sections 5.2, 5.3, and 7] The central claim that the discovered algorithms 'advance the state-of-the-art in Quality-Diversity optimization' is not supported by the paper's own numbers. Section 5.2 reports that LQD(F+N), the variant trained on the QD objective, 'approaches but does not quite match DNS's performance' and shows lower fitness than ME; Section 7 concedes that simultaneously optimizing both objectives remains challenging. The significant advantages are LQD(F) over GA on fitness (p<1e-10) and LQD(N) over Novelty Search on novelty (p<1e-9), but GA and NS are not QD baselines. On robot control, the aggregate significance claim in Section 5.3 explicitly excludes GA, and the QD-score comparison on out-of-distribution BBOB tasks in Figure 4 does not show LQD beating DNS on the QD score. Thus the paper's strongest framing is internally inconsistent with its reported evidence.
  2. [Section 4.1, Algorithm 3] The claim that LQD 'effectively searches through all archive mechanisms' is an overstatement that the architecture does not support. Algorithm 3 maintains a single fixed-size population and performs truncation selection on the learned competition scores; there is no persistent archive, no memory of previously seen individuals beyond the current population, and no grid or cell structure. Consequently the framework cannot represent MAP-Elites as actually implemented—where the archive persists across generations and each cell stores the best solution ever found—nor can it represent archive-based unstructured QD algorithms that maintain a separate collection of elites. The competition-function abstraction in Algorithm 2 places archive maintenance outside the learned component, so the paper's claim of searching over archive mechanisms is not justified.
  3. [Sections 2 and 5] The paper never compares LQD against the closest learned optimization baselines, LES (Lange et al., 2023, arXiv:2211.11260) and LGA (Lange et al., 2023, arXiv:2304.03995), despite discussing them as the most related work. Since LES and LGA are also attention-based, meta-trained meta-BBO algorithms, their absence makes it impossible to say whether LQD's improvements come from the QD formulation or from the general meta-BBO training scheme. At minimum, the paper should include these baselines on the shared BBOB and robot-control benchmarks, or explicitly state why they are not directly comparable.
minor comments (5)
  1. [Section 4.2.2] The QD score is defined as 'maximum fitness multipled by the average dominated novelty score'; 'multipled' should be 'multiplied', and the definition should clarify whether the maximum fitness and novelty are normalized and how they are aggregated across the population and generations.
  2. [Section 5.2 and Figure 3] The text states that LQD(F+N) outperforms ME in novelty with p<1e-7 but shows lower fitness; however, the figure legend lists 'MENS' as a baseline that is not defined anywhere in the text. This should be corrected to the full name (presumably ME-NS or Novelty Search with MAP-Elites) or removed.
  3. [Appendix B, Table 1] The transformer architecture is described as having '16 features per layer and 4 attention heads' and 'approximately 5,000 trainable parameters'; it would be helpful to state the number of parameters more precisely and to give the embedding dimension used for the descriptor-fitness featurization, since this affects the reproducibility of the architecture.
  4. [End of paper] The footer 'Received 29 January 2025; revised 12 March 2009; accepted 5 June 2009' appears to be a leftover template artifact and should be removed or replaced with the correct submission information.
  5. [Section 6.2 and Figure 8] The heatmaps in Figure 8 are described qualitatively but the color scale is not defined; adding a colorbar and a precise description of how the competition landscape is computed (e.g., which population snapshot and which median fitness value are used) would make the analysis reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the learned competition rules are meta-trained on a distinct distribution and then evaluated on held-out BBOB and robot-control tasks, so the central results do not reduce to their inputs.

full rationale

The derivation chain is self-contained rather than circular. The paper parameterizes the competition function with a transformer (Section 4.1) and meta-evolves its parameters on 22 BBOB functions with random-projection descriptors (Section 4.2.1), maximizing one of three explicit meta-objectives: 'Fitness: Optimizing the highest fitness value in the population', 'Novelty: Maximizing the average novelty score', or 'QD score: ... the maximum fitness multiplied by the average dominated novelty score' (Section 4.2.2). The headline results are not obtained by fitting to the evaluation targets: LQD(F) and LQD(N) are trained only on their respective single objectives, the out-of-distribution BBOB functions (Gallagher 101-me, Gallagher 21-hi, Ackley, Dixon-Price, Salomon, Levy) are excluded from training, and robot-control tasks involve higher dimensions, different descriptors, and a different simulator, so the generalization claim is externally tested rather than built into the objective. The self-citations to Lange et al. [29,30] for the meta-BBO procedure and to the authors' own DNS baseline [2] are not load-bearing: the meta-training procedure is reimplemented and empirically validated, and DNS is an independent comparison baseline rather than a premise used to derive LQD. No uniqueness theorem or ansatz is imported to force the architecture choice. Although the QD-score meta-objective is defined using dominated novelty, the corresponding variant does not match DNS, and the significant cross-domain results come from LQD(F) and LQD(N), so this benchmark alignment does not make the derivation circular. The paper's own limitation that 'LQD (F+N) approaches but does not consistently outperform existing QD algorithms' (Section 7) weakens the state-of-the-art framing, but that is a correctness and interpretation concern, not circularity. No fitted parameter is renamed as a prediction, and no evaluation metric is substituted into the training objective for the variants carrying the main claims.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central contribution is a learned parameterized competition function. Its weights and a set of hand-chosen meta-optimization hyperparameters are fitted to the training distribution; the paper does not release checkpoints or state the descriptor dimension. The main assumptions are the distance-preservation justification for random projection descriptors and the claim that QD algorithms can be captured by a stateless truncation-selection loop.

free parameters (4)
  • Learned LQD weights theta = Trained by Sep-CMA-ES on BBOB tasks; approx. 5,000 parameters; final values not released
    These weights are the discovered competition rule itself. They are fit to the meta-training distribution and are not derived from first principles.
  • Descriptor projection dimension D = Not specified in the paper
    The random matrix in Section 4.2.1 maps genotypes to descriptors, but the descriptor dimension used during meta-training is never stated. LQD performance likely depends on this choice.
  • Reproduction operator details = Not specified
    Algorithm 3 calls reproduction(X, f) without defining mutation rates, crossover, or offspring generation. Meta-training fixes this operator, so the discovered competition rules may co-adapt to it.
  • Meta-training hyperparameter set = N=128, B=32, T=256, M=256, K=256, 16,384 meta-generations, 4 layers, 16 features, 4 heads, sigma=0.1
    These hand-chosen hyperparameters from Table 1 shape the search space and the meta-optimization; no sensitivity analysis is provided.
assumptions (3)
  • standard math Johnson-Lindenstrauss lemma justifies random projection descriptors as meaningful for local competition
    Section 4.2.1 invokes the JL lemma to claim random projections approximately preserve pairwise distances. The distances preserved may not be the right notion of behavioral similarity for QD, and the transfer to hand-crafted descriptors is not theoretically grounded.
  • domain assumption QD algorithms can be abstracted as a genetic algorithm with local competition and truncation selection
    Algorithms 1 and 2 frame QD as a GA with top-N truncation. This excludes persistent archive structures such as MAP-Elites' grid archive, so the framework cannot represent all QD algorithms despite the claim in Section 4.1.
  • domain assumption Competition rules learned on random-projection descriptors on BBOB transfer to domain-specific descriptors in robot control
    Section 5.3 assumes the discovered rules generalize across descriptor distributions. This is empirically tested on 7 tasks but has no theoretical guarantee, and failure on other domains would invalidate the generalization claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Discovering Quality-Diversity Algorithms via Meta-Black-Box Optimization." pith.science (2026). https://pith.science/paper/6CONZLRU

@misc{pith2026250202190,
  author       = {Pith},
  title        = {Pith review of: Discovering Quality-Diversity Algorithms via Meta-Black-Box Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6CONZLRU}},
  note         = {Machine review of arXiv:2502.02190}
}
read the original abstract

Quality-Diversity has emerged as a powerful family of evolutionary algorithms that generate diverse populations of high-performing solutions by implementing local competition principles inspired by biological evolution. While these algorithms successfully foster diversity and innovation, their specific mechanisms rely on heuristics, such as grid-based competition in MAP-Elites or nearest-neighbor competition in unstructured archives. In this work, we propose a fundamentally different approach: using meta-learning to automatically discover novel Quality-Diversity algorithms. By parameterizing the competition rules using attention-based neural architectures, we evolve new algorithms that capture complex relationships between individuals in the descriptor space. Our discovered algorithms demonstrate competitive or superior performance compared to established Quality-Diversity baselines while exhibiting strong generalization to higher dimensions, larger populations, and out-of-distribution domains like robot control. Notably, even when optimized solely for fitness, these algorithms naturally maintain diverse populations, suggesting meta-learning rediscovers that diversity is fundamental to effective optimization.

Figures

Figures reproduced from arXiv: 2502.02190 by the authors.

Figure 1
Figure 1. Meta-black-box optimization of Learned Quality-Diversity (LQD) algorithms. At each meta-generation, a meta [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Learned competition function architecture. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Quality-Diversity trade-off across algorithms. Each [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (7 more)
Figure 5
Figure 5. Figure 5: Generalization analysis across population sizes and [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Fitness across robot control tasks. Lines show mean [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Novelty across robot control tasks. Lines show mean [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Impact of descriptors on optimization performance. [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 8
Figure 8. Figure 8: Visualization of learned competition landscapes [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 11
Figure 11. Figure 11: Impact of descriptors on optimization perfor [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 10
Figure 10. Figure 10: Performance comparison across meta-training BBOB tasks for the three distinct objectives. Results are normalized [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 46 canonical work pages

  1. [1]

    Marcin Andrychowicz, Misha Denil, Sergio Gomez, Matthew W Hoffman, David Pfau, Tom Schaul, Brendan Shillingford, and Nando De Freitas. 2016. Learning to learn by gradient descent by gradient descent. Advances in neural information processing systems 29 (2016). 8

    work page 2016

  2. [2]

    Ryan Bahlous-Boldi, Maxence Faldor, Luca Grillotti, Hannah Janmohamed, Lisa Coiffard, Lee Spector, and Antoine Cully. 2024. Dominated Novelty Search: Rethinking Local Competition in Quality-Diversity

    work page 2024

  3. [3]

    Samy Bengio, Yoshua Bengio, Jocelyn Cloutier, and Jan Gescei. 1992. On the optimization of a synaptic learning rule. In Optimality in Biological and Artificial Networks? Routledge, 281–303

    work page 1992

  4. [4]

    James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, and Qiao Zhang. 2018. JAX: composable transformations of Python+NumPy programs. http://github.com/jax-ml/jax

    2018

  5. [5]

    Felix Chalumeau, Bryan Lim, Raphaël Boige, Maxime Allard, Luca Grillotti, Manon Flageat, Valentin Macé, Guillaume Richard, Arthur Flajolet, Thomas Pierrot, and Antoine Cully. 2024. QDax: a library for quality-diversity and population-based algorithms with hardware acceleration. Journal of Machine Learning Research 25, 108 (2024), 1–16. http://jmlr.org/pap...

    work page 2024

  6. [6]

    Hoffman, Sergio Gómez Colmenarejo, Misha Denil, Timothy P

    Yutian Chen, Matthew W. Hoffman, Sergio Gómez Colmenarejo, Misha Denil, Timothy P. Lillicrap, Matt Botvinick, and Nando de Freitas. 2017. Learning to Learn without Gradient Descent by Gradient Descent. In Proceedings of the 34th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 70), Doina Precup and Yee Whye Teh ...

    work page 2017

  7. [7]

    Antoine Cully. 2019. Autonomous skill discovery with quality-diversity and unsupervised descriptors. In Proceedings of the Genetic and Evolutionary Com- putation Conference (GECCO ’19) . Association for Computing Machinery, New York, NY, USA, 81–89. doi:10.1145/3321707.3321804

    work page arXiv 2019

  8. [8]

    Antoine Cully, Jeff Clune, Danesh Tarapore, and Jean-Baptiste Mouret. 2015. Robots that can adapt like animals. Nature 521, 7553 (May 2015), 503–507. doi:10. 1038/nature14422

    work page 2015

Show all 55 references
  1. [9]

    Charles Darwin. 1859. On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life (1 ed.). John Murray, London

  2. [10]

    Maxence Faldor, Félix Chalumeau, Manon Flageat, and Antoine Cully. 2023. MAP-Elites with Descriptor-Conditioned Gradients and Archive Distillation into a Single Policy. In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO ’23). Association for Computing...

    2023

  3. [11]

    Maxence Faldor, Félix Chalumeau, Manon Flageat, and Antoine Cully. 2024. Syn- ergizing Quality-Diversity with Descriptor-Conditioned Reinforcement Learning. ACM Trans. Evol. Learn. Optim. (Sept. 2024). doi:10.1145/3696426

    2024 doi

  4. [12]

    Maxence Faldor and Antoine Cully. 2024. Toward Artificial Open-Ended Evolu- tion within Lenia using Quality-Diversity (Artificial Life Conference Proceed- ings, Vol. ALIFE 2024: Proceedings of the 2024 Artificial Life Conference) . 85. doi:10.1162/isal_a_00827

    2024 doi

  5. [13]

    Steffen Finck, Nikolaus Hansen, Raymond Ros, and Anne Auger. 2010. Real- Parameter Black-Box Optimization Benchmarking 2010: Presentation of the Noiseless Functions

    2010

  6. [14]

    Steffen Finck, Nikolaus Hansen, Raymond Ros, and Anne Auger. 2010. Real- Parameter Black-Box Optimization Benchmarking 2010: Presentation of the Noisy Functions

    2010

  7. [15]

    Daniel Freeman, Erik Frey, Anton Raichuk, Sertan Girgin, Igor Mordatch, and Olivier Bachem

    C. Daniel Freeman, Erik Frey, Anton Raichuk, Sertan Girgin, Igor Mordatch, and Olivier Bachem. 2021. Brax - A Differentiable Physics Engine for Large Scale Rigid Body Simulation. http://github.com/google/brax

    2021

  8. [16]

    Goldberg and Jon Richardson

    David E. Goldberg and Jon Richardson. 1987. Genetic Algorithms with Sharing for Multimodal Function Optimization. In Genetic Algorithms and their Applications: Proceedings of the Second International Conference on Genetic Algorithms . 41–49

    1987

  9. [17]

    Hugo Siqueira Gomes, Benjamin Léger, and Christian Gagné. 2021. Meta learning black-box population-based optimizers. arXiv preprint arXiv:2103.03526 (2021)

    2021 arXiv

  10. [18]

    Luca Grillotti and Antoine Cully. 2022. Unsupervised Behavior Discovery With Quality-Diversity Optimization. IEEE Transactions on Evolutionary Computation 26, 6 (2022), 1539–1552. doi:10.1109/TEVC.2022.3159855

    2022

  11. [19]

    Nikolaus Hansen and Andreas Ostermeier. 2001. Completely derandomized self-adaptation in evolution strategies. Evolutionary computation 9, 2 (2001), 159–195

    2001

  12. [20]

    John H. Holland. 1992. Genetic Algorithms. Scientific American 267, 1 (1992), 66–73. http://www.jstor.org/stable/24939139

    1992

  13. [21]

    Jianjun Hu, Erik Goodman, Kisung Seo, Zhun Fan, and Rondal Rosenberg. 2005. The Hierarchical Fair Competition (HFC) Framework for Sustainable Evolu- tionary Algorithms. Evol. Comput. 13, 2 (June 2005), 241–277. doi:10.1162/ 1063656054088530

    2005

  14. [22]

    Momin Jamil and Xin-She Yang. 2013. A Literature Survey of Benchmark Func- tions For Global Optimization Problems. Int. J. of Mathematical Modelling and Numerical Optimisation 4 (Aug. 2013). doi:10.1504/IJMMNO.2013.055204

    2013

  15. [23]

    Johnson and Joram Lindenstrauss

    William B. Johnson and Joram Lindenstrauss. 1984. Extensions of Lipschitz mappings into Hilbert space. Contemporary mathematics 26 (1984), 189–206. https://api.semanticscholar.org/CorpusID:117819162

    1984

  16. [24]

    Paul Kent, Juergen Branke, Adam Gaier, and Jean-Baptiste Mouret. 2022. A discretization-free metric for assessing quality diversity algorithms. InGECCO ’22: Proceedings of the Genetic and Evolutionary Computation Conference Companion . ACM, Massachusetts, Boston, 2131–2135. do...

    2022

  17. [25]

    Louis Kirsch, Sebastian Flennerhag, Hado van Hasselt, Abram Friesen, Junhyuk Oh, and Yutian Chen. 2022. Introducing symmetries to black box meta reinforce- ment learning. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 36. 7202–7210

    2022

  18. [26]

    Louis Kirsch and Jürgen Schmidhuber. 2021. Meta learning backpropagation and improving it. Advances in Neural Information Processing Systems 34 (2021), 14122–14134

    2021

  19. [27]

    Jannik Kossen, Neil Band, Clare Lyle, Aidan N Gomez, Thomas Rainforth, and Yarin Gal. 2021. Self-attention between datapoints: Going beyond individual input-output pairs in deep learning. Advances in Neural Information Processing Systems 34 (2021), 28742–28756

    2021

  20. [28]

    Robert Tjarko Lange. 2022. evosax: JAX-based Evolution Strategies.arXiv preprint arXiv:2212.04180 (2022)

    2022 arXiv

  21. [29]

    Robert Tjarko Lange, Tom Schaul, Yutian Chen, Chris Lu, Tom Zahavy, Valentin Dalibard, and Sebastian Flennerhag. 2023. Discovering Attention-Based Genetic Algorithms via Meta-Black-Box Optimization. doi:10.48550/arXiv.2304.03995 arXiv:2304.03995 [cs]

    2023 doi arXiv

  22. [30]

    Robert Tjarko Lange, Tom Schaul, Yutian Chen, Tom Zahavy, Valentin Dallibard, Chris Lu, Satinder Singh, and Sebastian Flennerhag. 2023. Discovering Evolu- tion Strategies via Meta-Black-Box Optimization. doi:10.48550/arXiv.2211.11260 arXiv:2211.11260 [cs]

    2023 doi arXiv

  23. [31]

    Robert Tjarko Lange, Yingtao Tian, and Yujin Tang. 2024. Evolution Trans- former: In-Context Evolutionary Optimization. doi:10.48550/arXiv.2403.02985 arXiv:2403.02985 [cs]

    2024 doi arXiv

  24. [32]

    Juho Lee, Yoonho Lee, Jungtaek Kim, Adam Kosiorek, Seungjin Choi, and Yee Whye Teh. 2019. Set transformer: A framework for attention-based permutation-invariant neural networks. In International conference on machine learning. PMLR, 3744–3753

    2019

  25. [33]

    Joel Lehman and Kenneth O. Stanley. 2011. Abandoning Objectives: Evolution Through the Search for Novelty Alone. Evolutionary Computation 19, 2 (June 2011), 189–223. doi:10.1162/EVCO_a_00025

    2011 doi

  26. [34]

    Joel Lehman and Kenneth O. Stanley. 2011. Evolving a diversity of virtual creatures through novelty search and local competition. InProceedings of the 13th Annual Conference on Genetic and Evolutionary Computation (Dublin, Ireland) (GECCO ’11). Association for Computing Machin...

    2011

  27. [35]

    Chris Lu, Jakub Grudzien Kuba, Alistair Letcher, Luke Metz, Christian Schroeder de Witt, and Jakob Nicolaus Foerster. 2022. Discovered Policy Optimisation. In Decision A wareness in Reinforcement Learning Workshop at ICML 2022

    2022

  28. [36]

    Luke Metz, Niru Maheswaranathan, Jeremy Nixon, Daniel Freeman, and Jascha Sohl-Dickstein. 2019. Understanding and correcting pathologies in the training of learned optimizers. In International Conference on Machine Learning . PMLR, 4556–4565

    2019

  29. [37]

    Jean-Baptiste Mouret and Jeff Clune. 2015. Illuminating search spaces by mapping elites. doi:10.48550/arXiv.1504.04909 arXiv:1504.04909 [cs]

    2015 doi arXiv

  30. [38]

    Junhyuk Oh, Matteo Hessel, Wojciech M Czarnecki, Zhongwen Xu, Hado P van Hasselt, Satinder Singh, and David Silver. 2020. Discovering reinforcement learning algorithms. Advances in Neural Information Processing Systems 33 (2020), 1060–1070

    2020

  31. [39]

    Jack Parker-Holder, Raghu Rajan, Xingyou Song, André Biedenkapp, Yingjie Miao, Theresa Eimer, Baohe Zhang, Vu Nguyen, Roberto Calandra, Aleksandra Faust, et al. 2022. Automated reinforcement learning (autorl): A survey and open problems. Journal of Artificial Intelligence Rese...

    2022

  32. [40]

    Pugh, Lisa B

    Justin K. Pugh, Lisa B. Soros, and Kenneth O. Stanley. 2016. Quality Diversity: A New Frontier for Evolutionary Computation. Frontiers in Robotics and AI 3 (July 2016). doi:10.3389/frobt.2016.00040 Publisher: Frontiers

    2016

  33. [41]

    Ingo Rechenberg and Manfred Eigen. 1973. Evolutionsstrategie: Optimierung technischer Systeme nach Prinzipien der biologischen Evolution . Number 15 in Problemata. Frommann-Holzboog, Stuttgart-Bad Cannstadt

    1973

  34. [42]

    Gresa Shala, André Biedenkapp, Noor Awad, Steven Adriaensen, Marius Lindauer, and Frank Hutter. 2020. Learning step-size adaptation in CMA-ES. InInternational Conference on Parallel Problem Solving from Nature . Springer, 691–706

    2020

  35. [43]

    Stanley and Joel Lehman

    Kenneth O. Stanley and Joel Lehman. 2015. Why Greatness Cannot Be Planned: The Myth of the Objective . Springer Publishing Company, Incorporated

    2015

  36. [44]

    Stanley and Risto Miikkulainen

    Kenneth O. Stanley and Risto Miikkulainen. 2002. Evolving Neural Networks Through Augmenting Topologies. Evolutionary Computation 10, 2 (2002), 99–127. http://nn.cs.utexas.edu/?stanley:ec02

    2002

  37. [45]

    Yujin Tang and David Ha. 2021. The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning. In Ad- vances in Neural Information Processing Systems , Vol. 34. Curran Associates, Inc., 22574–22587. https://proceedings.neurips.cc/paper_fi...

    2021

  38. [46]

    Vishnu TV, Pankaj Malhotra, Jyoti Narwariya, Lovekesh Vig, and Gautam Shroff

  39. [48]

    Vassilis Vassiliades, Konstantinos Chatzilygeroudis, and Jean-Baptiste Mouret

  40. [49]

    Gomez, Łukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Proceedings of the 31st International Conference on Neural Informa- tion Processing Systems (NIPS’17) . Curran Asso...

    2017

  41. [50]

    Jane X Wang, Zeb Kurth-Nelson, Dhruva Tirumala, Hubert Soyer, Joel Z Leibo, Remi Munos, Charles Blundell, Dharshan Kumaran, and Matt Botvinick. 2016. Learning to reinforcement learn. arXiv preprint arXiv:1611.05763 (2016)

    2016 arXiv

  42. [51]

    Zhongwen Xu, Hado P van Hasselt, Matteo Hessel, Junhyuk Oh, Satinder Singh, and David Silver. 2020. Meta-gradient reinforcement learning with an objective discovered online. Advances in Neural Information Processing Systems 33 (2020), 15254–15264

    2020

  43. [52]

    Zhongwen Xu, Hado P van Hasselt, and David Silver. 2018. Meta-gradient reinforcement learning. Advances in neural information processing systems 31 (2018)

    2018

  44. [53]

    Tom Zahavy, Zhongwen Xu, Vivek Veeriah, Matteo Hessel, Junhyuk Oh, Hado P van Hasselt, David Silver, and Satinder Singh. 2020. A self-tuning actor-critic algorithm. Advances in Neural Information Processing Systems 33 (2020), 20913– 20924. 10 Supplementary Materials A Addition...

    2020

  45. [2017]

    In Proceed- ings of the Genetic and Evolutionary Computation Conference Companion

    A comparison of illumination algorithms in unbounded spaces. In Proceed- ings of the Genetic and Evolutionary Computation Conference Companion . ACM, Berlin Germany, 1578–1581. doi:10.1145/3067695.3082531

  46. [2018]

    IEEE Transactions on Evolutionary Com- putation 22, 4 (Aug

    Using Centroidal Voronoi Tessellations to Scale Up the Multidimensional Archive of Phenotypic Elites Algorithm. IEEE Transactions on Evolutionary Com- putation 22, 4 (Aug. 2018), 623–630. doi:10.1109/TEVC.2017.2735550 Conference Name: IEEE Transactions on Evolutionary Computation

    2018

  47. [2019]

    In Joint European Conference on Machine Learning and Knowledge Discovery in Databases

    Meta-learning for black-box optimization. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases . Springer, 366–381. 9

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.