pith. sign in

arxiv: 2511.09598 · v5 · submitted 2025-11-12 · 💻 cs.LG

Amortized Multi-Objective Optimization Across Tasks with Generative Solution Modeling

Tingyang Wei , Jiao Liu , Abhishek Gupta , Chin Chun Ooi , Puay Siew Tan , Yew-Soon Ong This is my paper

Pith reviewed 2026-05-17 23:13 UTC · model grok-4.3

classification 💻 cs.LG
keywords multi-objective optimizationBayesian optimizationparametric problemsgenerative modelsinverse modelingamortized optimizationPareto solutionstask-aware Gaussian processes
0
0 comments X

The pith

Learning an inverse model allows direct prediction of Pareto solutions for new parametric multi-objective problems without re-evaluations.

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

The paper tackles families of expensive multi-objective optimization problems that vary continuously with task parameters, where solving each instance separately becomes prohibitive. It proposes learning an inverse model that maps any task and preference query straight to a set of solutions by alternating generative sampling of candidates with an acquisition search that exploits similarities across tasks. Task-aware Gaussian processes are used to justify faster convergence through shared structure. If the claim holds, handling an infinite continuum of such problems requires only a fixed upfront cost rather than repeated expensive runs. Empirical checks on synthetic and real-world cases are offered to support the approach.

Core claim

By alternating generative solution sampling via conditional generative models and acquisition-driven search with task-aware Gaussian processes, the parametric optimizer learns an inverse model that amortizes multi-objective optimization costs across the continuous task-preference space and enables accurate direct solution prediction for unseen parameterized expensive multi-objective problems without any further evaluations.

What carries the argument

The inverse model built by alternating conditional generative sampling of solutions and acquisition-driven search that leverages inter-task synergies.

If this is right

  • Inter-task synergies captured by task-aware Gaussian processes produce faster convergence than independent per-task runs.
  • After initial training, any new task in the parameter space receives a solution prediction at zero additional evaluation cost.
  • The method is shown to work on both synthetic benchmarks and real-world problem families where repeated evaluations would be costly.
  • Theoretical support is given for the convergence rate improvement stemming from shared task structure.

Where Pith is reading between the lines

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

  • The same amortization pattern could be tested on single-objective parametric problems or problems with constraints.
  • Selecting which initial tasks to evaluate could be guided by an active-learning criterion to maximize generalization of the inverse model.
  • The approach suggests a route to zero-shot querying in engineering design loops where operating conditions vary continuously.

Load-bearing premise

The inverse model trained on observed tasks generalizes accurately to new points in the continuous task parameter space without needing further evaluations.

What would settle it

Evaluate the model's predicted solution set for a held-out task parameter and compare it to the Pareto front obtained by running a standard optimizer from scratch on that same task; large dominance or quality gap would falsify the direct-prediction claim.

Figures

Figures reproduced from arXiv: 2511.09598 by Abhishek Gupta, Chin Chun Ooi, Jiao Liu, Puay Siew Tan, Tingyang Wei, Yew-Soon Ong.

Figure 1
Figure 1. Figure 1: The overall workflow of the proposed method. During the optimization process, solutions are sampled and [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Average hypervolume results evaluated on sampled EMOP tasks in synthetic and real-world benchmarks. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Approximated Pareto Front for unseen EMOP tasks in solar rooftop generation (top) and lamp generation [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Approximated Pareto Front for unseen EMOP tasks in solar rooftop generation (top) and lamp generation [PITH_FULL_IMAGE:figures/full_fig_p020_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Approximated Pareto Front for unseen EMOP tasks in solar rooftop generation (top) and lamp generation [PITH_FULL_IMAGE:figures/full_fig_p021_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Approximated Pareto Front for unseen EMOP tasks in solar rooftop generation (top) and lamp generation [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Approximated Pareto Front for unseen EMOP tasks in solar rooftop generation (top) and lamp generation [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Approximated Pareto Front for unseen EMOP tasks in solar rooftop generation (top) and lamp generation [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
read the original abstract

Many real-world applications require solving families of expensive multi-objective optimization problems~(EMOPs) under varying operational conditions. This can be formulated as parametric expensive multi-objective optimization problems (P-EMOPs) where each task parameter defines a distinct optimization instance. Current multi-objective Bayesian optimization methods have been widely used for finding finite sets of Pareto optimal solutions for each task. However, P-EMOPs present a fundamental challenge: the continuous task parameter space can contain infinite distinct problems, each requiring separate expensive evaluations. To address this, we propose learning an inverse model to amortize the multi-objective optimization cost across the continuous task-preference space, enabling direct solution prediction for any query without the need for expensive re-evaluation. This paper introduces a novel parametric multi-objective Bayesian optimizer that learns this inverse model by alternating between (1) generative solution sampling via conditional generative models and (2) acquisition-driven search leveraging inter-task synergies. This approach enables effective optimization across multiple tasks and finally achieves direct solution prediction for unseen parameterized EMOPs without re-evaluations. We theoretically justify the faster convergence by leveraging inter-task synergies through task-aware Gaussian processes. Based on that, empirical studies in synthetic and real-world benchmarks further verify the effectiveness of the proposed parametric optimizer.

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.

Referee Report

2 major / 2 minor

Summary. The paper proposes a parametric multi-objective Bayesian optimizer for families of expensive multi-objective optimization problems (P-EMOPs) parameterized by a continuous task variable. It learns an inverse model by alternating conditional generative sampling of solutions with acquisition-driven search that exploits inter-task synergies via task-aware Gaussian processes. The central claim is that this amortizes optimization across the task-preference space, enabling direct Pareto-set prediction for unseen tasks without further expensive evaluations. Theoretical justification for faster convergence is given, and empirical results on synthetic and real-world benchmarks are presented to support effectiveness.

Significance. If the generalization of the inverse model holds, the work would meaningfully advance amortized optimization for infinite families of EMOPs, offering substantial savings in evaluation budget for applications such as design under varying conditions or multi-scenario hyperparameter tuning. The integration of generative inverse modeling with task-aware GPs for synergy capture is a technically interesting direction.

major comments (2)
  1. [Abstract] Abstract: the claim that the method 'achieves direct solution prediction for unseen parameterized EMOPs without re-evaluations' is load-bearing for the contribution yet rests on unstated assumptions about the smoothness and learnability of the mapping from (task, preference) to solution sets; no generalization bounds, extrapolation analysis, or discussion of failure modes when objective landscapes vary non-smoothly with the task parameter are supplied.
  2. [Theoretical justification] Theoretical justification paragraph: the faster-convergence argument via inter-task synergies captured by task-aware GPs addresses sample efficiency during training but supplies no transfer guarantee or error bound for the conditional generative model's zero-shot accuracy on new task parameters, leaving the amortization benefit vulnerable to cases where predicted points are dominated or infeasible on the true objectives.
minor comments (2)
  1. Define all acronyms (EMOP, P-EMOP, GP) at first use and maintain consistent notation for task parameters and preference vectors throughout.
  2. [Experiments] In the experimental section, report the range and sampling density of task parameters used for training versus testing to allow assessment of extrapolation distance.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below and indicate the revisions that will be incorporated in the next version of the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that the method 'achieves direct solution prediction for unseen parameterized EMOPs without re-evaluations' is load-bearing for the contribution yet rests on unstated assumptions about the smoothness and learnability of the mapping from (task, preference) to solution sets; no generalization bounds, extrapolation analysis, or discussion of failure modes when objective landscapes vary non-smoothly with the task parameter are supplied.

    Authors: We agree that the abstract claim implicitly assumes sufficient smoothness and learnability of the mapping from (task, preference) pairs to Pareto sets. The current manuscript provides no formal generalization bounds. In the revision we will qualify the claim in the abstract, add a new paragraph in the Discussion section that explicitly states the smoothness assumption, reports the extrapolation behavior observed on the benchmarks, and discusses failure modes when objective landscapes change non-smoothly with the task parameter. revision: yes

  2. Referee: [Theoretical justification] Theoretical justification paragraph: the faster-convergence argument via inter-task synergies captured by task-aware GPs addresses sample efficiency during training but supplies no transfer guarantee or error bound for the conditional generative model's zero-shot accuracy on new task parameters, leaving the amortization benefit vulnerable to cases where predicted points are dominated or infeasible on the true objectives.

    Authors: The provided theoretical argument concerns only the sample-efficiency gains during the training phase that arise from inter-task synergies captured by the task-aware GPs. Zero-shot accuracy of the learned conditional generative model is supported empirically rather than by transfer bounds. We will revise the theoretical section to make this distinction explicit, add a short analysis of conditions under which predicted points may become dominated or infeasible, and include additional diagnostic plots from the existing benchmarks that illustrate robustness and degradation cases. revision: partial

Circularity Check

0 steps flagged

No circularity; derivation introduces new amortization via generative inverse model and task-aware GPs

full rationale

The paper proposes a novel parametric multi-objective Bayesian optimizer that learns an inverse model by alternating generative solution sampling via conditional generative models and acquisition-driven search with task-aware Gaussian processes to leverage inter-task synergies. It claims this enables direct solution prediction for unseen P-EMOPs without re-evaluations and provides a theoretical justification for faster convergence. No equations, fitted parameters renamed as predictions, self-definitional reductions, or load-bearing self-citations appear in the abstract or description. The central claim rests on the construction of the new alternating procedure and generalization of the learned model rather than presupposing the target result or reducing predictions to inputs by construction. The derivation is therefore self-contained against external benchmarks such as standard MOBO and generative modeling techniques.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only access prevents extraction of specific free parameters, axioms, or invented entities; the method appears to build on standard Bayesian optimization and generative modeling without introducing new postulated entities in the summary.

pith-pipeline@v0.9.0 · 5535 in / 1078 out tokens · 35521 ms · 2026-05-17T23:13:45.684602+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

62 extracted references · 62 canonical work pages

  1. [1]

    J. Knowles. Parego: a hybrid algorithm with on-line landscape approximation for expensive multiobjective optimization problems.IEEE Trans. on Evol. Comput., 10(1):50–66, 2006

    work page 2006

  2. [2]

    Multiproblem surrogates: Transfer evolutionary multiobjective optimization of computationally expensive problems.IEEE Trans.on Evol

    Alan Tan Wei Min, Yew-Soon Ong, Abhishek Gupta, and Chi-Keong Goh. Multiproblem surrogates: Transfer evolutionary multiobjective optimization of computationally expensive problems.IEEE Trans.on Evol. Comput., 23(1):15–28, 2019

    work page 2019

  3. [3]

    Schoenholz, Muratahan Aykol, Gowoon Cheon, and Ekin Dogus Cubuk

    Amil Merchant, Simon Batzner, Samuel S. Schoenholz, Muratahan Aykol, Gowoon Cheon, and Ekin Dogus Cubuk. Scaling deep learning for materials discovery.Nature, 624(7990):80–85, Dec 2023

    work page 2023

  4. [4]

    Computational discovery of microstructured composites with optimal stiffness-toughness trade-offs.Science Advances, 10(5):eadk4284, 2024

    Beichen Li, Bolei Deng, Wan Shou, Tae-Hyun Oh, Yuanming Hu, Yiyue Luo, Liang Shi, and Wojciech Matusik. Computational discovery of microstructured composites with optimal stiffness-toughness trade-offs.Science Advances, 10(5):eadk4284, 2024

    work page 2024

  5. [5]

    Offline model-based optimization for real-world molecular discovery

    Dong-Hee Shin, Young-Han Son, Hyun Jung Lee, Deok-Joong Lee, and Tae-Eui Kam. Offline model-based optimization for real-world molecular discovery. InForty-second International Conference on Machine Learning, 2025

    work page 2025

  6. [6]

    Sample-efficient multi-objective molecular optimization with gflownets

    Yiheng Zhu, Jialu Wu, Chaowen Hu, Jiahuan Yan, kim hsieh, Tingjun Hou, and Jian Wu. Sample-efficient multi-objective molecular optimization with gflownets. InAdvances in Neural Information Processing Systems, volume 36, pages 79667–79684. Curran Associates, Inc., 2023

    work page 2023

  7. [7]

    Robust model-free reinforcement learning with multi- objective bayesian optimization

    Matteo Turchetta, Andreas Krause, and Sebastian Trimpe. Robust model-free reinforcement learning with multi- objective bayesian optimization. In2020 IEEE International Conference on Robotics and Automation (ICRA), pages 10702–10708, 2020

    work page 2020

  8. [8]

    Mo-bbo: Multi-objective bilevel bayesian optimization for robot and behavior co-design

    Yeonju Kim, Zherong Pan, and Kris Hauser. Mo-bbo: Multi-objective bilevel bayesian optimization for robot and behavior co-design. In2021 IEEE International Conference on Robotics and Automation (ICRA), pages 9877–9883, 2021

    work page 2021

  9. [9]

    A flexible framework for multi-objective bayesian optimization using random scalarizations

    Biswajit Paria, Kirthevasan Kandasamy, and Barnabás Póczos. A flexible framework for multi-objective bayesian optimization using random scalarizations. InProceedings of The 35th Uncertainty in Artificial Intelligence Conference, volume 115 ofProceedings of Machine Learning Research, pages 766–776. PMLR, 22–25 Jul 2020

    work page 2020

  10. [10]

    Emmerich, K.C

    M.T.M. Emmerich, K.C. Giannakoglou, and B. Naujoks. Single- and multiobjective evolutionary optimization assisted by gaussian random field metamodels.IEEE Trans. on Evol. Comput., 10(4):421–439, 2006. 10 APREPRINT- JANUARY12, 2026

    work page 2006

  11. [11]

    Differentiable expected hypervolume improvement for parallel multi-objective bayesian optimization.Advances in Neural Information Processing Systems, 33:9851– 9864, 2020

    Samuel Daulton, Maximilian Balandat, and Eytan Bakshy. Differentiable expected hypervolume improvement for parallel multi-objective bayesian optimization.Advances in Neural Information Processing Systems, 33:9851– 9864, 2020

    work page 2020

  12. [12]

    Multiobjective optimization using gaussian process emulators via stepwise uncertainty reduction

    Victor Picheny. Multiobjective optimization using gaussian process emulators via stepwise uncertainty reduction. Statistics and Computing, 25(6):1265–1280, 2015

    work page 2015

  13. [13]

    Predictive entropy search for multi-objective bayesian optimization

    Daniel Hernandez-Lobato, Jose Hernandez-Lobato, Amar Shah, and Ryan Adams. Predictive entropy search for multi-objective bayesian optimization. InProceedings of The 33rd International Conference on Machine Learning, volume 48 ofProceedings of Machine Learning Research, pages 1492–1501, New York, New York, USA, 20–22 Jun 2016. PMLR

    work page 2016

  14. [14]

    Joint entropy search for multi-objective bayesian optimization

    Ben Tu, Axel Gandy, Nikolas Kantas, and Behrang Shafei. Joint entropy search for multi-objective bayesian optimization. InAdvances in Neural Information Processing Systems, volume 35, pages 9922–9938. Curran Associates, Inc., 2022

    work page 2022

  15. [15]

    Max-value entropy search for multi-objective bayesian optimization

    Syrine Belakaria, Aryan Deshwal, and Janardhan Rao Doppa. Max-value entropy search for multi-objective bayesian optimization. InAdvances in Neural Information Processing Systems, volume 32. Curran Associates, Inc., 2019

    work page 2019

  16. [16]

    Pareto set learning for expensive multi-objective optimization

    Xi Lin, Zhiyuan Yang, Xiaoyuan Zhang, and Qingfu Zhang. Pareto set learning for expensive multi-objective optimization. InAdvances in Neural Information Processing Systems, volume 35, pages 19231–19247. Curran Associates, Inc., 2022

    work page 2022

  17. [17]

    Expensive multi-objective bayesian optimization based on diffusion models.Proceedings of the AAAI Conference on Artificial Intelligence, 39(25):27063–27071, Apr

    Bingdong Li, Zixiang Di, Yongfan Lu, Hong Qian, Feng Wang, Peng Yang, Ke Tang, and Aimin Zhou. Expensive multi-objective bayesian optimization based on diffusion models.Proceedings of the AAAI Conference on Artificial Intelligence, 39(25):27063–27071, Apr. 2025

    work page 2025

  18. [18]

    Parametric pareto set learning for expensive multi-objective optimization, 2025

    Ji Cheng, Bo Xue, and Qingfu Zhang. Parametric pareto set learning for expensive multi-objective optimization, 2025

    work page 2025

  19. [19]

    Pareto gamuts: exploring optimal designs across varying contexts.ACM Trans

    Liane Makatura, Minghao Guo, Adriana Schulz, Justin Solomon, and Wojciech Matusik. Pareto gamuts: exploring optimal designs across varying contexts.ACM Trans. on Graph., 40(4), July 2021

    work page 2021

  20. [20]

    Robots that can adapt like animals.Nature, 521(7553):503–507, May 2015

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

    work page 2015

  21. [21]

    Anna Nickelson, Kagan Tumer, and William D. Smart. Contextual multi-objective path planning. In2023 IEEE International Conference on Robotics and Automation (ICRA), pages 10240–10246, 2023

    work page 2023

  22. [22]

    Springer Science & Business Media, 2008

    Jürgen Branke.Multiobjective optimization: Interactive and evolutionary approaches, volume 5252. Springer Science & Business Media, 2008

    work page 2008

  23. [23]

    Multi-objective quality diversity optimiza- tion

    Thomas Pierrot, Guillaume Richard, Karim Beguir, and Antoine Cully. Multi-objective quality diversity optimiza- tion. InProceedings of the Genetic and Evolutionary Computation Conference, GECCO ’22, page 139–147, New York, NY , USA, 2022. Association for Computing Machinery

    work page 2022

  24. [24]

    Springer Science & Business Media, 1999

    Kaisa Miettinen.Nonlinear multiobjective optimization, volume 12. Springer Science & Business Media, 1999

    work page 1999

  25. [25]

    A review on evolutionary multitask optimization: Trends and challenges.IEEE Trans

    Tingyang Wei, Shibin Wang, Jinghui Zhong, Dong Liu, and Jun Zhang. A review on evolutionary multitask optimization: Trends and challenges.IEEE Trans. on Evol. Comput., 26(5):941–960, 2022

    work page 2022

  26. [26]

    Multiobjective multifactorial optimization in evolutionary multitasking.IEEE Trans

    Abhishek Gupta, Yew-Soon Ong, Liang Feng, and Kay Chen Tan. Multiobjective multifactorial optimization in evolutionary multitasking.IEEE Trans. on Cybern., 47(7):1652–1665, 2016

    work page 2016

  27. [27]

    Multifactorial evolution: toward evolutionary multitasking

    Abhishek Gupta, Yew-Soon Ong, and Liang Feng. Multifactorial evolution: toward evolutionary multitasking. IEEE Trans. on Evol. Comput., 20(3):343–357, 2016

    work page 2016

  28. [28]

    Bayesian forward-inverse transfer for multiobjective optimization

    Tingyang Wei, Jiao Liu, Abhishek Gupta, Puay Siew Tan, and Yew-Soon Ong. Bayesian forward-inverse transfer for multiobjective optimization. InParallel Problem Solving from Nature – PPSN XVIII, pages 135–152, Cham,

    work page

  29. [29]

    Springer Nature Switzerland

    work page

  30. [30]

    Cognizant multitasking in multiob- jective multifactorial evolution: Mo-mfea-ii.IEEE Trans

    Kavitesh Kumar Bali, Abhishek Gupta, Yew-Soon Ong, and Puay Siew Tan. Cognizant multitasking in multiob- jective multifactorial evolution: Mo-mfea-ii.IEEE Trans. on Cybern., 51(4):1784–1796, 2021

    work page 2021

  31. [31]

    Parametric pareto set learning: Amortizing multi-objective optimization with parameters.IEEE Transactions on Evolutionary Computation, pages 1–1, 2025

    Ji Cheng, Xi Lin, Bo Xue, and Qingfu Zhang. Parametric pareto set learning: Amortizing multi-objective optimization with parameters.IEEE Transactions on Evolutionary Computation, pages 1–1, 2025

    work page 2025

  32. [32]

    C. K. Williams and C. E. Rasmussen.Gaussian Processes for Machine Learning. MIT Press, 2006

    work page 2006

  33. [33]

    Kakade, and Matthias W

    Niranjan Srinivas, Andreas Krause, Sham M. Kakade, and Matthias W. Seeger. Information-theoretic regret bounds for Gaussian process optimization in the bandit setting.IEEE Trans. on Information Theory, 58(5):3250–3265, 2012. 11 APREPRINT- JANUARY12, 2026

    work page 2012

  34. [34]

    Random hypervolume scalarizations for provable multi-objective black box optimization

    Richard Zhang and Daniel Golovin. Random hypervolume scalarizations for provable multi-objective black box optimization. InProceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 11096–11105. PMLR, 13–18 Jul 2020

    work page 2020

  35. [35]

    Conditioning by adaptive sampling for robust design

    David Brookes, Hahnbeom Park, and Jennifer Listgarten. Conditioning by adaptive sampling for robust design. In Proceedings of the 36th International Conference on Machine Learning, volume 97 ofProceedings of Machine Learning Research, pages 773–782. PMLR, 09–15 Jun 2019

    work page 2019

  36. [36]

    Unifying likelihood-free inference with black-box sequence design and beyond

    Dinghuai Zhang, Jie Fu, Yoshua Bengio, and Aaron Courville. Unifying likelihood-free inference with black-box sequence design and beyond. InInternational Conference on Learning Representations, 2022

    work page 2022

  37. [37]

    Auto-encoding variational bayes, 2022

    Diederik P Kingma and Max Welling. Auto-encoding variational bayes, 2022

    work page 2022

  38. [38]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. InAdvances in Neural Information Processing Systems, volume 33, pages 6840–6851. Curran Associates, Inc., 2020

    work page 2020

  39. [39]

    Learning structured output representation using deep conditional generative models

    Kihyuk Sohn, Honglak Lee, and Xinchen Yan. Learning structured output representation using deep conditional generative models. InAdvances in Neural Information Processing Systems, volume 28. Curran Associates, Inc., 2015

    work page 2015

  40. [40]

    Multiobjective tree-structured parzen estimator.J

    Yoshihiko Ozaki, Yuki Tanigaki, Shuhei Watanabe, Masahiro Nomura, and Masaki Onishi. Multiobjective tree-structured parzen estimator.J. Artif. Int. Res., 73, May 2022

    work page 2022

  41. [41]

    Multiobjective optimization using evolutionary algorithms — a comparative case study

    Eckart Zitzler and Lothar Thiele. Multiobjective optimization using evolutionary algorithms — a comparative case study. InParallel Problem Solving from Nature — PPSN V, pages 292–301, Berlin, Heidelberg, 1998. Springer Berlin Heidelberg

    work page 1998

  42. [42]

    K. Deb, L. Thiele, M. Laumanns, and E. Zitzler. Scalable multi-objective optimization test problems. In Proceedings of the 2002 Congress on Evolutionary Computation. CEC’02, volume 1, pages 825–830 vol.1, 2002

    work page 2002

  43. [43]

    Extremo: Transfer evolutionary multiobjective optimization with proof of faster convergence.IEEE Trans

    Jiao Liu, Abhishek Gupta, Chinchun Ooi, and Yew-Soon Ong. Extremo: Transfer evolutionary multiobjective optimization with proof of faster convergence.IEEE Trans. on Evol. Comput., 29(1):102–116, 2025

    work page 2025

  44. [44]

    Underactuated robotics: Algorithms for walking, running, swimming, flying, and manipulation

    Russ Tedrake. Underactuated robotics: Algorithms for walking, running, swimming, flying, and manipulation. 2018.URL http://underactuated. mit. edu, 2022

    work page 2018

  45. [45]

    Botorch: A framework for efficient monte-carlo bayesian optimization

    Maximilian Balandat, Brian Karrer, Daniel Jiang, Samuel Daulton, Ben Letham, Andrew G Wilson, and Eytan Bakshy. Botorch: A framework for efficient monte-carlo bayesian optimization. InAdvances in Neural Information Processing Systems, volume 33, pages 21524–21538. Curran Associates, Inc., 2020

    work page 2020

  46. [46]

    Offline contextual Bayesian optimization

    Ian Char, Youngseog Chung, Willie Neiswanger, Kirthevasan Kandasamy, Andrew Oakleigh Nelson, Mark Boyer, Egemen Kolemen, and Jeff Schneider. Offline contextual Bayesian optimization. InAdvances in Neural Information Processing Systems, volume 32. Curran Associates, Inc., 2019

    work page 2019

  47. [47]

    High dimensional bayesian optimisation and bandits via additive models

    Kirthevasan Kandasamy, Jeff Schneider, and Barnabas Poczos. High dimensional bayesian optimisation and bandits via additive models. InProceedings of the 32nd International Conference on Machine Learning, volume 37 ofProceedings of Machine Learning Research, pages 295–304, Lille, France, 07–09 Jul 2015. PMLR

    work page 2015

  48. [48]

    Ioannis Giagkiozis and Peter J. Fleming. Pareto front estimation for decision making.Evolutionary Computation, 22(4):651–678, 12 2014

    work page 2014

  49. [49]

    A multiobjective evolutionary algorithm using gaussian process-based inverse modeling.IEEE Transactions on Evolutionary Computation, 19(6):838–856, 2015

    Ran Cheng, Yaochu Jin, Kaname Narukawa, and Bernhard Sendhoff. A multiobjective evolutionary algorithm using gaussian process-based inverse modeling.IEEE Transactions on Evolutionary Computation, 19(6):838–856, 2015

    work page 2015

  50. [50]

    Inverse gaussian process modeling for evolutionary dynamic multiobjective optimization.IEEE Transactions on Cybernetics, 52(10):11240–11253, 2022

    Huan Zhang, Jinliang Ding, Min Jiang, Kay Chen Tan, and Tianyou Chai. Inverse gaussian process modeling for evolutionary dynamic multiobjective optimization.IEEE Transactions on Cybernetics, 52(10):11240–11253, 2022

    work page 2022

  51. [51]

    Pareto optimization with small data by learning across common objective spaces.Scientific Reports, 13(1):7842, May 2023

    Chin Sheng Tan, Abhishek Gupta, Yew-Soon Ong, Mahardhika Pratama, Puay Siew Tan, and Siew Kei Lam. Pareto optimization with small data by learning across common objective spaces.Scientific Reports, 13(1):7842, May 2023

    work page 2023

  52. [52]

    Multi-task bayesian optimization.Advances in neural information processing systems, 26, 2013

    Kevin Swersky, Jasper Snoek, and Ryan P Adams. Multi-task bayesian optimization.Advances in neural information processing systems, 26, 2013

    work page 2013

  53. [53]

    Few-shot bayesian optimization with deep kernel surrogates, 2021

    Martin Wistuba and Josif Grabocka. Few-shot bayesian optimization with deep kernel surrogates, 2021

    work page 2021

  54. [54]

    Reinforced few-shot acquisition function learning for bayesian optimization

    Bing-Jing Hsieh, Ping-Chun Hsieh, and Xi Liu. Reinforced few-shot acquisition function learning for bayesian optimization. InAdvances in Neural Information Processing Systems, volume 34, pages 7718–7731. Curran Associates, Inc., 2021. 12 APREPRINT- JANUARY12, 2026

    work page 2021

  55. [55]

    Dahl, Kevin Swersky, Chansoo Lee, Zachary Nado, Justin Gilmer, Jasper Snoek, and Zoubin Ghahramani

    Zi Wang, George E. Dahl, Kevin Swersky, Chansoo Lee, Zachary Nado, Justin Gilmer, Jasper Snoek, and Zoubin Ghahramani. Pre-trained gaussian processes for bayesian optimization.Journal of Machine Learning Research, 25(212):1–83, 2024

    work page 2024

  56. [56]

    Transfer learning for bayesian optimization on heterogeneous search spaces

    Zhou Fan, Xinran Han, and Zi Wang. Transfer learning for bayesian optimization on heterogeneous search spaces. Transactions on Machine Learning Research, 2024

    work page 2024

  57. [57]

    Regret Bounds for Transfer Learning in Bayesian Optimisation

    Alistair Shilton, Sunil Gupta, Santu Rana, and Svetha Venkatesh. Regret Bounds for Transfer Learning in Bayesian Optimisation. InProceedings of the 20th International Conference on Artificial Intelligence and Statistics, volume 54 ofProceedings of Machine Learning Research, pages 307–315. PMLR, 20–22 Apr 2017

    work page 2017

  58. [58]

    Springer Science & Business Media, 2006

    Fuzhen Zhang.The Schur complement and its applications, volume 4. Springer Science & Business Media, 2006

    work page 2006

  59. [59]

    Diffusion models for black-box optimiza- tion

    Siddarth Krishnamoorthy, Satvik Mehul Mashkaria, and Aditya Grover. Diffusion models for black-box optimiza- tion. InProceedings of the 40th International Conference on Machine Learning, volume 202 ofProceedings of Machine Learning Research, pages 17842–17857. PMLR, 23–29 Jul 2023

    work page 2023

  60. [60]

    Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps

    Cheng Lu, Yuhao Zhou, Fan Bao, Jianfei Chen, Chongxuan LI, and Jun Zhu. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. InAdvances in Neural Information Processing Systems, volume 35, pages 5775–5787. Curran Associates, Inc., 2022

    work page 2022

  61. [61]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization, 2017

    work page 2017

  62. [62]

    get lucky

    Jacob Gardner, Geoff Pleiss, Kilian Q Weinberger, David Bindel, and Andrew G Wilson. Gpytorch: Blackbox matrix-matrix gaussian process inference with gpu acceleration. InAdvances in Neural Information Processing Systems, volume 31. Curran Associates, Inc., 2018. Appendix A Proofs of Theorem 2 Proof.For taskk, let γT,m,k = max {x(t) k }T t=1⊂X I {y(t) m,k}...

    work page 2018