arxiv: 2604.21991 · v1 · submitted 2026-04-23 · 💻 cs.LG · cs.AI· cs.NE

Recognition: unknown

Multi-Task Optimization over Networks of Tasks

Julian Hatzky , Thomas Bartz-Beielstein , A. E. Eiben , Anil Yaman

Authors on Pith no claims yet

Pith reviewed 2026-05-09 22:52 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.NE

keywords multi-task optimizationtask networksknowledge transferMAP-Elitesevolutionary algorithmsgraph-based searchcontinuous control

0 comments

The pith

MONET models tasks as a connected graph so that crossover from parameter-space neighbors transfers solutions while mutation refines each task independently.

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

The paper introduces MONET to overcome scaling limits in multi-task optimization. Population-based methods struggle with large task sets, and existing scalable alternatives rely on fixed discretized archives that ignore how tasks relate in parameter space. MONET builds an explicit graph with tasks as nodes and edges linking nearby tasks, then alternates social learning (crossover from neighbors) with individual learning (mutation on a node's own solution). Experiments on four domains, each with thousands of tasks, show performance that meets or beats MAP-Elites baselines. A reader cares because the method keeps the task space continuous and topology-aware rather than forcing a rigid grid.

Core claim

By representing the task space as a graph whose edges connect tasks according to their parameter-space proximity, MONET enables knowledge transfer through neighbor crossover while each task still undergoes independent mutation, producing results that match or exceed those of MAP-Elites-based methods on archery, arm, cartpole, and hexapod domains containing 2,000 to 5,000 tasks.

What carries the argument

The task network graph, whose edges connect tasks in parameter space so that crossover can move solutions between neighboring nodes.

If this is right

The approach remains tractable for high-dimensional task spaces because it never builds an explicit archive grid.
Knowledge transfer occurs only between tasks that are close in parameter space, preserving locality.
Social and individual learning can be interleaved at each generation without requiring global population maintenance.
The same graph construction works across continuous control and locomotion tasks without domain-specific tuning.

Where Pith is reading between the lines

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

If parameter-space proximity fails to predict solution similarity in a new domain, the graph edges would need to be replaced by a learned similarity measure.
The method could be extended to dynamically add or remove tasks by updating only the local neighborhood rather than rebuilding a global archive.
Because the graph is explicit, it may support theoretical analysis of information flow rates between tasks that fixed-archive methods do not.

Load-bearing premise

Connecting tasks by their raw parameter-space distance creates useful opportunities for solution transfer via crossover.

What would settle it

Run MONET on one of the four domains after replacing the parameter-space edges with random edges; if performance drops below the MAP-Elites baseline, the topology assumption is falsified.

Figures

Figures reproduced from arXiv: 2604.21991 by A. E. Eiben, Anil Yaman, Julian Hatzky, Thomas Bartz-Beielstein.

**Figure 2.** Figure 2: Commutative diagrams comparing MAP-Elites and MONET. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: Overview of benchmark environments. To assess statistical significance of the performance differences between algorithms, we apply the Mann-Whitney U test [35] to the final fitness values and AUC values across all seeds and report the resulting p-values with HolmBonferroni correction [20, 26] for multiple comparisons. Higher fitness is better in all four benchmarks, but the scale differs across them: arc… view at source ↗

**Figure 4.** Figure 4: Algorithm comparison across the four benchmark domains. [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗

**Figure 5.** Figure 5: Low values of pind (favoring social learning over individual mutation) consistently improve both mean fitness and AUC across all domains, supporting the central design intuition behind MONET that information transfer between similar tasks is the dominant lever. The remaining hyperparameters show weaker and more domain-dependent effects (see SI for detailed SHAP analysis per domain [25]) [PITH_FULL_IMAGE:… view at source ↗

**Figure 5.** Figure 5: Global hyperparameter importance for MONET: mean absolute SHAP [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗

**Figure 6.** Figure 6: SHAP beeswarm plot for the archery domain. Each row is a hyperparam [PITH_FULL_IMAGE:figures/full_fig_p023_6.png] view at source ↗

**Figure 7.** Figure 7: SHAP beeswarm plot for the arm domain. Panels as in Figure 6. [PITH_FULL_IMAGE:figures/full_fig_p024_7.png] view at source ↗

**Figure 8.** Figure 8: SHAP beeswarm plot for the cartpole domain. Panels as in Figure 6. [PITH_FULL_IMAGE:figures/full_fig_p025_8.png] view at source ↗

**Figure 9.** Figure 9: SHAP beeswarm plot for the hexapod domain. Panels as in Figure 6. [PITH_FULL_IMAGE:figures/full_fig_p026_9.png] view at source ↗

read the original abstract

Multi-task optimization is a powerful approach for solving a large number of tasks in parallel. However, existing algorithms face distinct limitations: Population-based methods scale poorly and remain underexplored for large task sets. Approaches that do scale beyond a thousand tasks are mostly MAP-Elites variants and rely on a fixed, discretized archive that disregards the topology of the task space. We introduce MONET (Multi-Task Optimization over Networks of Tasks), a multi-task optimization algorithm that models the task space as a graph: tasks are nodes, and edges connect tasks in the task parameter space. This representation enables knowledge transfer between tasks and remains tractable for high-dimensional problems while exploiting the topology of the task space. MONET combines social learning, which generates candidates from neighboring nodes via crossover, with individual learning, which refines a node's own solution independently via mutation. We evaluate MONET on four domains (archery, arm, and cartpole with 5,000 tasks each; hexapod with 2,000 tasks) and show that it matches or exceeds the performance of existing MAP-Elites-based baselines across all four domains.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

MONET adds a parameter-space graph plus neighbor crossover to multi-task evolutionary optimization, but the abstract gives no ablation or numbers to show the graph itself matters.

read the letter

MONET treats tasks as nodes in a graph with edges drawn from parameter similarity, then runs crossover between neighbors for social learning and mutation for individual refinement. This is the main new piece relative to fixed-archive MAP-Elites methods that ignore task topology. The setup is evaluated on four domains with 2,000–5,000 tasks each, and the abstract says it matches or exceeds the baselines. If the full paper supplies reproducible code, clear metrics, and statistical tests, that would be a practical data point for people scaling population methods in robotics and control. The experiments look aimed at exactly the scaling regime where grid-based archives become awkward. The main gap is the missing ablation: nothing in the abstract checks whether random edges or pure individual learning would produce the same outcome. Without that isolation, any gains could come from the dual learning loop rather than the topology modeling that is supposed to be the distinguishing feature. The abstract also omits quantitative results, error bars, or baseline implementation details, which leaves the performance claim hard to judge. This paper is for readers already working on evolutionary multi-task methods who want to see a graph-based alternative tried at scale. It is coherent on its own terms and engages the right literature, so it deserves a serious referee to examine the full results and any ablations that may be present.

Referee Report

2 major / 1 minor

Summary. The paper introduces MONET, a multi-task optimization algorithm that models the task space as a graph with tasks as nodes and edges connecting tasks in parameter space. This enables knowledge transfer via social learning (crossover with neighboring nodes) combined with individual learning (mutation). The authors evaluate MONET on four domains (archery, arm, and cartpole with 5,000 tasks each; hexapod with 2,000 tasks) and claim it matches or exceeds the performance of existing MAP-Elites-based baselines across all four domains.

Significance. If the empirical claims are substantiated with detailed metrics and controls, this work could advance scalable multi-task optimization by replacing fixed discretized archives with a topology-exploiting graph representation, addressing scalability limits of population-based methods for large task sets in domains like robotics.

major comments (2)

[Abstract] Abstract: The claim that MONET 'matches or exceeds the performance of existing MAP-Elites-based baselines across all four domains' supplies no quantitative metrics, error bars, statistical tests, or baseline implementation details. This is load-bearing for the central empirical contribution and prevents verification of whether the result holds.
[Method] Method description: No ablation is reported that replaces parameter-space neighbor selection with random edges (or removes social learning) while holding the dual learning loop fixed. Without this isolation, performance gains cannot be attributed specifically to the task graph topology rather than general social+individual learning.

minor comments (1)

[Abstract] Abstract: The domain list uses ambiguous shorthand ('arm') and provides no details on how the task parameter space is defined or how edges are constructed in the graph.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which will help improve the clarity and rigor of our work. We address the major comments point by point below.

read point-by-point responses

Referee: [Abstract] Abstract: The claim that MONET 'matches or exceeds the performance of existing MAP-Elites-based baselines across all four domains' supplies no quantitative metrics, error bars, statistical tests, or baseline implementation details. This is load-bearing for the central empirical contribution and prevents verification of whether the result holds.

Authors: We concur that the abstract would benefit from more quantitative backing to support the central claim. In the revised manuscript, we will modify the abstract to incorporate key quantitative metrics from our experiments, including average performance scores with standard deviations, and mention that statistical tests were conducted to compare against baselines. We will also briefly reference the implementation details of the MAP-Elites baselines used in the evaluation. revision: yes
Referee: [Method] Method description: No ablation is reported that replaces parameter-space neighbor selection with random edges (or removes social learning) while holding the dual learning loop fixed. Without this isolation, performance gains cannot be attributed specifically to the task graph topology rather than general social+individual learning.

Authors: The referee raises an important point regarding causal attribution. Our primary results compare MONET to established MAP-Elites methods, which lack the graph-based social learning mechanism. To further isolate the role of the task graph, we will include an additional ablation experiment in the revised paper. This will involve running MONET with random edge connections instead of parameter-space neighbors, while maintaining the crossover and mutation operations, to demonstrate that the topology-aware neighbor selection contributes to the observed performance. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical algorithm evaluation is self-contained

full rationale

The paper defines MONET explicitly as a graph-based multi-task optimizer that connects tasks in parameter space and combines neighbor crossover (social learning) with per-node mutation (individual learning). The headline result is an empirical comparison showing MONET matches or exceeds MAP-Elites baselines on four fixed domains. No equations, uniqueness theorems, or first-principles derivations appear in the provided text; the task-graph topology is an input design choice rather than a quantity predicted from the algorithm itself. No fitted parameters are renamed as predictions, and no load-bearing step reduces to a self-citation chain. The central claim therefore rests on external experimental outcomes rather than internal tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Review based on abstract only; no explicit free parameters, axioms, or invented entities are detailed beyond the high-level graph representation.

axioms (1)

domain assumption Tasks possess a topology in parameter space that can be represented as a graph with edges between similar tasks.
This assumption underpins the knowledge-transfer mechanism via neighbor crossover.

invented entities (1)

MONET algorithm no independent evidence
purpose: Graph-based multi-task optimizer using social and individual learning.
New method introduced in the paper.

pith-pipeline@v0.9.0 · 5505 in / 1123 out tokens · 35061 ms · 2026-05-09T22:52:09.248795+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

55 extracted references · 36 canonical work pages · 2 internal anchors

[1]

Alba, E., Dorronsoro, B.: Cellular Genetic Algorithms, Operations Re- search/Computer Science Interfaces Series, vol. 42. Springer US, Boston, MA (2008).https://doi.org/10.1007/978-0-387-77610-1

work page doi:10.1007/978-0-387-77610-1 2008
[2]

Proximal Policy Optimization Algorithms

and Filip Wolski and Prafulla Dhariwal and Alec Radford and Oleg Klimov, J.S.: Proximal policy optimization algorithms. CoRRabs/1707.06347(2017)

work page internal anchor Pith review Pith/arXiv arXiv 2017
[3]

In: Proceedings of the Companion Conference on Genetic and Evolutionary Computation

Anne, T., Mouret, J.B.: Multi-task multi-behavior MAP-elites. In: Proceedings of the Companion Conference on Genetic and Evolutionary Computation. pp. 111–
[4]

GECCO ’23 Companion, Association for Computing Machinery, New York, NY, USA (2023).https://doi.org/10.1145/3583133.3590730

work page doi:10.1145/3583133.3590730 2023
[5]

In: Proceedings of the Ge- netic and Evolutionary Computation Conference

Anne, T., Mouret, J.B.: Parametric-Task MAP-Elites. In: Proceedings of the Ge- netic and Evolutionary Computation Conference. pp. 68–77 (Jul 2024).https: //doi.org/10.1145/3638529.3653993

work page doi:10.1145/3638529.3653993 2024
[6]

Anne, T., Mouret, J.B.: Parametric-Task MAP-Elites.https://github.com/ hucebot/Parametric-Task_MAP-Elites(2024)

2024
[7]

IEEE Transactions on Cybernetics 51(4), 1784–1796 (2021).https://doi.org/10.1109/TCYB.2020.2981733

Bali, K.K., Gupta, A., Ong, Y.S., Tan, P.S.: Cognizant multitasking in multiob- jective multifactorial evolution: MO-MFEA-II. IEEE Transactions on Cybernetics 51(4), 1784–1796 (2021).https://doi.org/10.1109/TCYB.2020.2981733

work page doi:10.1109/tcyb.2020.2981733 2021
[8]

IEEE Transactions on Evolu- tionaryComputation24(1),69–83(2020).https://doi.org/10.1109/TEVC.2019

Bali, K.K., Ong, Y.S., Gupta, A., Tan, P.S.: Multifactorial evolutionary algorithm with online transfer parameter estimation: MFEA-II. IEEE Transactions on Evolu- tionaryComputation24(1),69–83(2020).https://doi.org/10.1109/TEVC.2019. 2906927

work page doi:10.1109/tevc.2019 2020
[9]

IEEE Transactions on Systems, Man, and CyberneticsSMC-13(5), 834–846 (1983).https://doi.org/10.1109/TSMC

Barto, A.G., Sutton, R.S., Anderson, C.W.: Neuronlike adaptive elements that can solve difficult learning control problems. IEEE Transactions on Systems, Man, and CyberneticsSMC-13(5), 834–846 (1983).https://doi.org/10.1109/TSMC. 1983.6313077

work page doi:10.1109/tsmc 1983
[10]

Bartz-Beielstein, T.: Optimization with spotoptim (2026),https://arxiv.org/ abs/2604.13672

work page internal anchor Pith review Pith/arXiv arXiv 2026
[11]

Communications of the ACM18(9), 509–517 (1975).https://doi.org/10.1145/ 361002.361007

Bentley, J.L.: Multidimensional binary search trees used for associative searching. Communications of the ACM18(9), 509–517 (1975).https://doi.org/10.1145/ 361002.361007

work page arXiv 1975
[12]

Science314(5802), 1118–1121 (2006).https://doi.org/10.1126/ science.1133687

Bongard, J., Zykov, V., Lipson, H.: Resilient machines through continuous self-modeling. Science314(5802), 1118–1121 (2006).https://doi.org/10.1126/ science.1133687

2006
[13]

In: Black Box Optimiza- tion, Machine Learning, and No-Free Lunch Theorems, pp

Chatzilygeroudis, K., Cully, A., Vassiliades, V., Mouret, J.B.: Quality-diversity optimization: A novel branch of stochastic optimization. In: Black Box Optimiza- tion, Machine Learning, and No-Free Lunch Theorems, pp. 109–135. Springer Optimization and Its Applications, Springer (2021).https://doi.org/10.1007/ 978-3-030-66515-9_4

2021
[14]

Coumans, E., Bai, Y.: PyBullet, a Python module for physics simulation for games, robotics and machine learning.http://pybullet.org(2021)

2021
[15]

Robots that can adapt like animals

Cully, A., Clune, J., Tarapore, D., Mouret, J.B.: Robots that can adapt like animals. Nature521(7553), 503–507 (2015).https://doi.org/10.1038/ nature14422,https://doi.org/10.1038/nature14422

work page doi:10.1038/nature14422 2015
[16]

Wiley, Chichester, UK (2001)

Deb, K.: Multi-Objective Optimization Using Evolutionary Algorithms. Wiley, Chichester, UK (2001)

2001
[17]

Complex Systems9(2), 115–148 (1995) 16 J

Deb, K., Agrawal, R.B.: Simulated binary crossover for continuous search space. Complex Systems9(2), 115–148 (1995) 16 J. Hatzky et al

1995
[18]

In: Proceedings of the Mathematics, Toronto

Delaunay, B.: Sur la sphere vide. In: Proceedings of the Mathematics, Toronto. pp. 695–700. Toronto (1928)

1928
[19]

Nature Communications15, 6267 (2024).https://doi.org/10.1038/s41467-024-50131-4

van Diggelen, F., Cambier, N., Ferrante, E., Eiben, A.E.: A model-free method to learn multiple skills in parallel on modular robots. Nature Communications15, 6267 (2024).https://doi.org/10.1038/s41467-024-50131-4

work page doi:10.1038/s41467-024-50131-4 2024
[20]

SIAM Review41(4), 637–676 (1999).https://doi.org/10.1137/ S0036144599352836

Du, Q., Faber, V., Gunzburger, M.: Centroidal voronoi tessellations: Applications and algorithms. SIAM Review41(4), 637–676 (1999).https://doi.org/10.1137/ S0036144599352836

1999
[21]

Journal of the American Sta- tistical Association56(293), 52–64 (1961).https://doi.org/10.1080/01621459

Dunn, O.J.: Multiple comparisons among means. Journal of the American Sta- tistical Association56(293), 52–64 (1961).https://doi.org/10.1080/01621459. 1961.10482090

work page doi:10.1080/01621459 1961
[22]

Feller, W.: An Introduction to Probability Theory and Its Applications, vol. 1. Wiley, 3 edn. (1968)

1968
[23]

Feng, L., Huang, Y., Zhou, L., Zhong, J., Gupta, A., Tang, K., Tan, K.C.: Explicit evolutionary multitasking for combinatorial optimization: A case study on capaci- tatedvehicleroutingproblem.IEEETransactionsonCybernetics51(6),3143–3156 (2021).https://doi.org/10.1109/TCYB.2019.2962865

work page doi:10.1109/tcyb.2019.2962865 2021
[24]

IEEE Transactions on Evolutionary Computation20(3), 343–357 (2016)

Gupta, A., Ong, Y.S., Feng, L.: Multifactorial evolution: Toward evolutionary mul- titasking. IEEE Transactions on Evolutionary Computation20(3), 343–357 (2016). https://doi.org/10.1109/TEVC.2015.2458037

work page doi:10.1109/tevc.2015.2458037 2016
[25]

IEEE Computational Intelligence Magazine17, 49–66 (May 2022).https://doi.org/10.1109/MCI.2022.3155332

Gupta, A., Zhou, L., Ong, Y.S., Chen, Z., Hou, Y.: Half a dozen real-world appli- cations of evolutionary multitasking, and more. IEEE Computational Intelligence Magazine17, 49–66 (May 2022).https://doi.org/10.1109/MCI.2022.3155332

work page doi:10.1109/mci.2022.3155332 2022
[26]

Hatzky, J., Bartz-Beielstein, T., Eiben, A.E., Yaman, A.: Repository for MONET: Multi-Task Optimization over Networks of Tasks.https://github.com/ju2ez/ MONET/tree/PPSN-MONET-introduction(2026)

2026
[27]

Scandinavian Journal of Statistics6(2), 65–70 (1979)

Holm, S.: A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics6(2), 65–70 (1979)

1979
[28]

IEEE Transactions on Systems, Man, and Cybernetics: Systems55(5), 3492–3505 (May 2025).https://doi.org/10.1109/TSMC.2025.3541002

Hou, Y., Yu, Z., Guo, Z., Pei, W., Wu, Y., Ge, H., Xue, B., Zhang, M.: A Group- Based Many-Task Collaborative Optimization Framework for Evolutionary Robots Design. IEEE Transactions on Systems, Man, and Cybernetics: Systems55(5), 3492–3505 (May 2025).https://doi.org/10.1109/TSMC.2025.3541002

work page doi:10.1109/tsmc.2025.3541002 2025
[29]

In: Proceedings of the 31st International Conference on Machine Learning

Hutter, F., Hoos, H., Leyton-Brown, K.: An efficient approach for assessing hy- perparameter importance. In: Proceedings of the 31st International Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 32, pp. 754–
[30]

IEEE Transactions on Evolutionary Computation27(6), 1735–1749 (Dec 2023).https://doi.org/10.1109/TEVC.2022.3227120

Huynh Thi Thanh, B., Van Cuong, L., Thang, T.B., Long, N.H.: Ensemble Mul- tifactorial Evolution With Biased Skill-Factor Inheritance for Many-Task Opti- mization. IEEE Transactions on Evolutionary Computation27(6), 1735–1749 (Dec 2023).https://doi.org/10.1109/TEVC.2022.3227120

work page doi:10.1109/tevc.2022.3227120 2023
[31]

Energy Conversion and Management207, 112509 (Mar 2020).https://doi.org/ 10.1016/j.enconman.2020.112509

Liang, J., Qiao, K., Yuan, M., Yu, K., Qu, B., Ge, S., Li, Y., Chen, G.: Evolu- tionary multi-task optimization for parameters extraction of photovoltaic models. Energy Conversion and Management207, 112509 (Mar 2020).https://doi.org/ 10.1016/j.enconman.2020.112509

work page doi:10.1016/j.enconman.2020.112509 2020
[32]

https://doi.org/10.1109/ACCESS.2020.3018484

Liu, J., Li, P., Wang, G., Zha, Y., Peng, J., Xu, G.: A multitasking electric power dispatch approach with multi-objective multifactorial optimization algo- rithm.IEEEaccess:practicalinnovations,opensolutions8,155902–155911(2020). https://doi.org/10.1109/ACCESS.2020.3018484

work page doi:10.1109/access.2020.3018484 2020
[33]

In: 2022 IEEE Congress on Evolutionary Computation (CEC)

Liu, P., Guo, Z., Yu, H., Linghu, H., Li, Y., Hou, Y., Ge, H., Zhang, Q.: A pre- liminary study of multi-task map-elites with knowledge transfer for robotic arm Multi-Task Optimization over Networks of Tasks 17 design. In: 2022 IEEE Congress on Evolutionary Computation (CEC). pp. 1–8 (2022).https://doi.org/10.1109/CEC55065.2022.9870374

work page doi:10.1109/cec55065.2022.9870374 2022
[34]

Andrea Cristina McGlinchey and Peter J

Lundberg, S.M., Erion, G., Chen, H., DeGrave, A., Prutkin, J.M., Nair, B., Katz, R., Himmelfarb, J., Bansal, N., Lee, S.I.: From local explanations to global under- standing with explainable AI for trees. Nature Machine Intelligence2(1), 56–67 (2020).https://doi.org/10.1038/s42256-019-0138-9

work page doi:10.1038/s42256-019-0138-9 2020
[35]

In: Advances in Neural Information Processing Systems

Lundberg, S.M., Lee, S.I.: A unified approach to interpreting model predictions. In: Advances in Neural Information Processing Systems. vol. 30 (2017)

2017
[36]

The Annals of Mathematical Statistics , author =

Mann, H.B., Whitney, D.R.: On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics18(1), 50–60 (1947).https://doi.org/10.1214/aoms/1177730491

work page doi:10.1214/aoms/1177730491 1947
[37]

In: Dale, E., Michie, D

Michie, D., Chambers, R.A.: BOXES: An experiment in adaptive control. In: Dale, E., Michie, D. (eds.) Machine Intelligence 2, pp. 137–152. Oliver and Boyd, Edin- burgh (1968)

1968
[38]

Mouret, J.B.: pymap_elites: Python reference implementation of MAP-Elites and Multi-Task MAP-Elites.https://github.com/resibots/pymap_elites
[39]

Illuminating search spaces by mapping elites

Mouret, J., Clune, J.: Illuminating search spaces by mapping elites. CoRR abs/1504.04909(2015),http://arxiv.org/abs/1504.04909

work page Pith review arXiv 2015
[40]

In: Pro- ceedings of the 2020 Genetic and Evolutionary Computation Conference

Mouret, J.B., Maguire, G.: Quality diversity for multi-task optimization. In: Pro- ceedings of the 2020 Genetic and Evolutionary Computation Conference. pp. 121–129. ACM, Cancún Mexico (Jun 2020).https://doi.org/10.1145/3377930. 3390203

work page doi:10.1145/3377930 2020
[41]

Osaba, E., Martinez, A.D., Ser, J.D.: Evolutionary multitask optimization: a methodological overview, challenges and future research directions (2021),https: //arxiv.org/abs/2102.02558

work page arXiv 2021
[42]

European Journal of Operational Research270(3), 1074–1085 (2018)

Pearce, M., Branke, J.: Continuous multi-task bayesian optimisation with cor- relation. European Journal of Operational Research270(3), 1074–1085 (2018). https://doi.org/10.1016/j.ejor.2018.03.017

work page doi:10.1016/j.ejor.2018.03.017 2018
[43]

Science328(5975), 208–213 (2010)

Rendell, L., Boyd, R., Cownden, D., Enquist, M., Eriksson, K., Feldman, M.W., Fogarty, L., Ghirlanda, S., Lillicrap, T., Laland, K.N.: Why copy others? Insights from the social learning strategies tournament. Science328(5975), 208–213 (2010). https://doi.org/10.1126/science.1184719

work page doi:10.1126/science.1184719 2010
[44]

In: 2016 IEEE Symposium Series on Computa- tional Intelligence (SSCI)

Sagarna, R., Ong, Y.S.: Concurrently searching branches in software tests genera- tion through multitask evolution. In: 2016 IEEE Symposium Series on Computa- tional Intelligence (SSCI). pp. 1–8 (2016).https://doi.org/10.1109/SSCI.2016. 7850040

work page doi:10.1109/ssci.2016 2016
[45]

Expert Systems with Applications201, 117060 (2022).https://doi.org/ 10.1016/j.eswa.2022.117060

Samarakoon, S.M.B.P., Muthugala, M.A.V.J., Elara, M.R.: Metaheuristic based navigation of a reconfigurable robot through narrow spaces with shape changing ability. Expert Systems with Applications201, 117060 (2022).https://doi.org/ 10.1016/j.eswa.2022.117060

work page doi:10.1016/j.eswa.2022.117060 2022
[46]

doi: 10.1109/JPROC.2015.2494218

Shahriari, B., Swersky, K., Wang, Z., Adams, R.P., de Freitas, N.: Taking the human out of the loop: A review of bayesian optimization. Proceedings of the IEEE104(1), 148–175 (2016).https://doi.org/10.1109/JPROC.2015.2494218

work page doi:10.1109/jproc.2015.2494218 2016
[47]

In: Proceedings of the Genetic and Evolutionary Com- putation Conference

Triebold, C., Yaman, A.: Evolving generalist controllers to handle a wide range of morphological variations. In: Proceedings of the Genetic and Evolutionary Com- putation Conference. pp. 1137–1145 (2024)

2024
[48]

Which of the following [...]

Vassiliades,V.,Mouret,J.B.:Discoveringtheelitehypervolumebyleveraginginter- species correlation. In: Proceedings of the Genetic and Evolutionary Computation Conference. pp. 149–156. Gecco ’18, Association for Computing Machinery, New York, NY, USA (2018).https://doi.org/10.1145/3205455.3205602 18 J. Hatzky et al

work page doi:10.1145/3205455.3205602 2018
[49]

IEEE Transactions on Evolutionary ComputationPP, 1–1 (Aug 2017)

Vassiliades, V., Chatzilygeroudis, K., Mouret, J.B.: Using centroidal voronoi tes- sellations to scale up the multi-dimensional archive of phenotypic elites algo- rithm. IEEE Transactions on Evolutionary ComputationPP, 1–1 (Aug 2017). https://doi.org/10.1109/TEVC.2017.2735550

work page doi:10.1109/tevc.2017.2735550 2017
[50]

Neural networks regularization with graph- based local resampling

Wang, C., Liu, J., Wu, K., Wu, Z.: Solving Multitask Optimization Problems With Adaptive Knowledge Transfer via Anomaly Detection. IEEE Transactions on Evo- lutionary Computation26(2), 304–318 (Apr 2022).https://doi.org/10.1109/ TEVC.2021.3068157

work page arXiv 2022
[51]

PLoS computational biology18(2), e1009882 (2022)

Yaman, A., Bredeche, N., Çaylak, O., Leibo, J.Z., Lee, S.W.: Meta-control of social learning strategies. PLoS computational biology18(2), e1009882 (2022)

2022
[52]

Applied Soft Computing101, 106993 (2021)

Yaman, A., Iacca, G.: Distributed embodied evolution over networks. Applied Soft Computing101, 106993 (2021)

2021
[53]

IEEE Robotics & Au- tomation Magazine14(1), 43–52 (2007).https://doi.org/10.1109/MRA.2007

Yim, M., Shen, W.M., Salemi, B., Rus, D., Moll, M., Lipson, H., Klavins, E., Chirikjian, G.S.: Modular self-reconfigurable robot systems. IEEE Robotics & Au- tomation Magazine14(1), 43–52 (2007).https://doi.org/10.1109/MRA.2007. 339623

work page doi:10.1109/mra.2007 2007
[54]

Applied Soft Computing145, 110545 (2023).https://doi.org/10.1016/j.asoc.2023.110545

Zhao, H., Ning, X., Liu, X., Wang, C., Liu, J.: What makes evolutionary multi- task optimization better: A comprehensive survey. Applied Soft Computing145, 110545 (2023).https://doi.org/10.1016/j.asoc.2023.110545

work page doi:10.1016/j.asoc.2023.110545 2023
[55]

IEEE Transactions on Systems, Man, and Cyber- netics: Systems50(11), 4492–4505 (2020).https://doi.org/10.1109/TSMC.2018

Zhong, J., Feng, L., Cai, W., Ong, Y.S.: Multifactorial genetic programming for symbolic regression problems. IEEE Transactions on Systems, Man, and Cyber- netics: Systems50(11), 4492–4505 (2020).https://doi.org/10.1109/TSMC.2018. 2853719 A Node Coverage Guarantees Under Random Sampling If we draw at each evaluation step a random node and perform an actio...

work page doi:10.1109/tsmc.2018 2020