Mapping pareto fronts for efficient multi-objective materials discovery

Abstract

With advancements in automation and high-throughput techniques, we can tackle more complex multi-objective materials discovery problems requiring a higher evaluation budget. Given that experimentation is greatly limited by evaluation budget, maximizing sample efficiency of optimization becomes crucial. We discuss the limitations of using hypervolume as a performance indicator and propose new metrics relevant to materials experimentation: such as the ability to perform well for complex high-dimensional problems, minimizing wastage of evaluations, consistency/robustness of optimization, and ability to scale well to high throughputs. With these metrics, we perform an empirical study of two conceptually different and state-of-the-art algorithms (Bayesian and Evolutionary) on synthetic and real-world datasets. We discuss the merits of both approaches with respect to exploration and exploitation, where fully resolving the Pareto Front provides more knowledge of the best material.

Keywords

Bayesian optimization / constrained multi-objective optimization / evolutionary algorithm / materials science

References

[1] Correa-Baena J,van Duren J.Accelerating materials development via automation, machine learning, and high-performance computing.Joule2018;2:1410-20

[2] Mennen SM,Allen CL.The evolution of high-throughput experimentation in pharmaceutical development and perspectives on the future.Org Process Res Dev2019;23:1213-42

[3] Burger B,Gusev VV.A mobile robotic chemist.Nature2020;583:237-41

[4] Langner S,Perea JD.Beyond ternary OPV: high-throughput experimentation and self-driving laboratories optimize multicomponent systems.Adv Mater2020;32:e1907801

[5] Bash D,Ren Z.Accelerated automated screening of viscous graphene suspensions with various surfactants for optimal electrical conductivity.Digit Discov2022;1:139-46

[6] Mekki-Berrada F,Huang T.Two-step machine learning enables optimized nanoparticle synthesis.NPJ Comput Mater2021;7

[7] Li Z,Deng Y,Tasan CC.Metastable high-entropy dual-phase alloys overcome the strength-ductility trade-off.Nature2016;534:227-30

[8] Ramirez I,Zhong Y,Riede M.Key tradeoffs limiting the performance of organic photovoltaics.Adv Energy Mater2018;8:1703551

[9] Ren S,Salvatore D.Molecular electrocatalysts can mediate fast, selective CO₂ reduction in a flow cell.Science2019;365:367-9

[10] Bash D,Chellappan V.Machine learning and high-throughput robust design of P3HT-CNT composite thin films for high electrical conductivity.arXiv preprint2020;[Accepted]:2011.10382

[11] Grizou J,Sharma A.A curious formulation robot enables the discovery of a novel protocell behavior.Sci Adv2020;6:eaay4237 PMCID:PMC6994213

[12] Abdel-Latif K,Bateni F,Reyes KG.Self-driven multistep quantum dot synthesis enabled by autonomous robotic experimentation in flow.Adv Intell Syst2021;3:2000245

[13] Yong W,Fu H,He J.Improving prediction accuracy of high-performance materials via modified machine learning strategy.Comput Mater Sci2022;204:111181

[14] Lim YF,Vaitesswar U.Extrapolative bayesian optimization with gaussian process and neural network ensemble surrogate models.Adv Intell Syst2021;3:2100101

[15] Klein A,Falkner S,Hutter F. Towards efficient Bayesian optimization for big data. Available from: https://bayesopt.github.io/papers/2015/klein.pdf [Last accessed on 21 Apr 2023]

[16] Gopakumar AM,Xue D,Lookman T.Multi-objective optimization for materials discovery via adaptive design.Sci Rep2018;8:3738 PMCID:PMC5829239

[17] Niculescu RS,Rao RB,Parrado-Hernández E. Bayesian network learning with parameter constraints. Available from: https://www.jmlr.org/papers/volume7/niculescu06a/niculescu06a.pdf [Last accessed on 21 Apr 2023]

[18] Asvatourian V,Michiels S.Integrating expert’s knowledge constraint of time dependent exposures in structure learning for Bayesian networks.Artif Intell Med2020;107:101874

[19] Liu Z,Flick AC.Machine learning with knowledge constraints for process optimization of open-air perovskite solar cell manufacturing.Joule2022;6:834-49

[20] Deb K.Multi-objective optimisation using evolutionary algorithms: an introduction. In: Multi-Objective Evolutionary Optimisation for Product Design and Manufacturing. Springer; 2011:3-34.

[21] Mitchell M.An introduction to genetic algorithms. MIT press; 1998.

[22] Fan Z,Cai X.An improved epsilon constraint-handling method in MOEA/D for CMOPs with large infeasible regions.Soft Comput2019;23:12491-510

[23] Xu B.Constrained optimization based on ensemble differential evolution and two-level-based epsilon method.IEEE Access2020;8:213981-97

[24] Tian Y,Su Y,Tan KC.Balancing objective optimization and constraint satisfaction in constrained evolutionary multiobjective optimization.IEEE Trans Cybern2022;52:9559-72

[25] Li B,Tang K.Many-objective evolutionary algorithms: a survey.ACM Comput Surv2015;48:1-35

[26] Zhang P,Qian Q.Multi-objective optimization for materials design with improved NSGA-II.Mater Today Commun2021;28:102709

[27] Jha R,Dulikravich GS,Chakraborti N.Evolutionary design of nickel-based superalloys using data-driven genetic algorithms and related strategies.Mater Manuf Process2015;30:488-510

[28] Coello CA, Becerra RL. Evolutionary multiobjective optimization in materials science and engineering.Mater Manuf Process2009;24:119-29

[29] Pakhnova M,Yanilkin A.Search for stable cocrystals of energetic materials using the evolutionary algorithm USPEX.Phys Chem Chem Phys2020;22:16822-30

[30] Jennings PC,Hummelshøj JS,Bligaard T.Genetic algorithms for computational materials discovery accelerated by machine learning.NPJ Comput Mater2019;5

[31] Salley D,Grizou J,Martín S.A nanomaterials discovery robot for the Darwinian evolution of shape programmable gold nanoparticles.Nat Commun2020;11:2771 PMCID:PMC7265452

[32] Shahriari B,Wang Z,de Freitas N.Taking the human out of the loop: a review of Bayesian optimization.Proc IEEE2016;104:148-75

[33] Rasmussen CE.Gaussian processes in machine learning. In: Lecture notes in computer science. Springer; 2004;3176:63-71.

[34] Garrido-merchán EC.Predictive entropy search for multi-objective Bayesian optimization with constraints.Neurocomputing2019;361:50-68

[35] Fernández-Sánchez D,Hernández-Lobato D.Max-value entropy search for multi-objective bayesian optimization with constraints.arXiv preprint2020:2009.01721

[36] Suzuki S,Tamura T,Karasuyama M. Multi-objective Bayesian optimization using pareto-frontier entropy. Available from: http://proceedings.mlr.press/v119/suzuki20a/suzuki20a.pdf [Last accessed on 21 Apr 2023]

[37] Bradford E,Lapkin A.Efficient multiobjective optimization employing Gaussian processes, spectral sampling and a genetic algorithm.J Glob Optim2018;71:407-38

[38] Zhang, Wudong Liu, Tsang E, Virginas B. Expensive multiobjective optimization by MOEA/D with gaussian process model.IEEE Trans Evol Computat2010;14:456-74

[39] Mannodi-Kanakkithodi A,Ramprasad R,Gubernatis JE.Multi-objective optimization techniques to design the Pareto front of organic dielectric polymers.Comput Mater Sci2016;125:92-9

[40] Yuan R,Balachandran PV.Accelerated discovery of large electrostrains in BaTiO₃ -based piezoelectrics using active learning.Adv Mater2018;30:1702884

[41] Karasuyama M,Tamura T.Computational design of stable and highly ion-conductive materials using multi-objective bayesian optimization: case studies on diffusion of oxygen and lithium.Comput Mater Sci2020;184:109927

[42] MacLeod BP,Rupnow CC.A self-driving laboratory advances the Pareto front for material properties.Nat Commun2022;13:995 PMCID:PMC8863835

[43] Cao L,Felton K.Optimization of formulations using robotic experiments driven by machine learning DoE.Cell Rep Phys Sci2021;2:100295

[44] Erps T,Luković MK.Accelerated discovery of 3D printing materials using data-driven multiobjective optimization.Sci Adv2021;7:eabf7435 PMCID:PMC8519564

[45] Epps RW,Volk AA.Artificial chemist: an autonomous quantum dot synthesis bot.Adv Mater2020;32:e2001626

[46] Hanaoka K.Comparison of conceptually different multi-objective Bayesian optimization methods for material design problems.Mater Today Commun2022;31:103440

[47] Auger A,Brockhoff D.Hypervolume-based multiobjective optimization: Theoretical foundations and practical implications.Theor Comput Sci2012;425:75-103

[48] Guerreiro AP,Paquete L.The hypervolume indicator: problems and algorithms.arXiv preprint2020;[Accepted]:

[49] Häse F,Aspuru-Guzik A.Chimera: enabling hierarchy based multi-objective optimization for self-driving laboratories.Chem Sci2018;9:7642-55 PMCID:PMC6182568

[50] Zhang H,Zhu S,Xie J.Machine learning assisted composition effective design for precipitation strengthened copper alloys.Acta Mater2021;215:117118

[51] Daulton S,Bakshy E. Parallel bayesian optimization of multiple noisy objectives with expected hypervolume improvement. Available from: https://proceedings.neurips.cc/paper/2021/file/11704817e347269b7254e744b5e22dac-Paper.pdf [Last accessed on 21 Apr 2023]

[52] Seada H.U-NSGA-III: a unified evolutionary optimization procedure for single, multiple, and many objectives: proof-of-principle results. In: Gaspar-cunha A, Henggeler Antunes C, Coello CC, editors. Evolutionary multi-criterion optimization. Cham: Springer International Publishing; 2015. pp. 34-49.

[53] Jones DR.A taxonomy of global optimization methods based on response surfaces.J Glob Optim21:345-83

[54] Daulton S,Bakshy E. Differentiable expected hypervolume improvement for parallel multi-objective Bayesian optimization. Available from: https://proceedings.neurips.cc/paper/2020/file/6fec24eac8f18ed793f5eaad3dd7977c-Paper.pdf [Last accessed on 21 Apr 2023]

[55] Balandat M,Jiang D. BoTorch: a framework for efficient Monte-Carlo Bayesian optimization. Available from: https://proceedings.neurips.cc/paper/2020/file/f5b1b89d98b7286673128a5fb112cb9a-Paper.pdf [Last accessed on 21 Apr 2023]

[56] Deb K.An evolutionary many-objective optimization algorithm using reference-point-based nondominated sorting approach, part I: solving problems with box constraints.IEEE Trans Evol Computat2014;18:577-601

[57] Jain H.An evolutionary many-objective optimization algorithm using reference-point based nondominated sorting approach, Part II: handling constraints and extending to an adaptive approach.IEEE Trans Evol Computat2014;18:602-22

[58] Blank J.Pymoo: multi-objective optimization in python.IEEE Access2020;8:89497-509

[59] Zitzler E,Thiele L.Comparison of multiobjective evolutionary algorithms: empirical results.Evol Comput2000;8:173-95

[60] Ma Z.Evolutionary constrained multiobjective optimization: test suite construction and performance comparisons.IEEE Trans Evol Computat2019;23:972-86

[61] MacLeod BP,Morrissey TD.Self-driving laboratory for accelerated discovery of thin-film materials.Sci Adv2020;6:eaaz8867 PMCID:PMC7220369

[62] Yeh IC. Modeling slump of concrete with fly ash and superplasticizer. Available from: https://www.dbpia.co.kr/Journal/articleDetail?nodeId=NODE10903242 [Last accessed on 21 Apr 2023]

[63] Moriconi R,Sesh Kumar KS.High-dimensional Bayesian optimization using low-dimensional feature spaces.Mach Learn2020;109:1925-43

[64] Eriksson D. High-dimensional Bayesian optimization with sparse axis-aligned subspaces. Available from: https://proceedings.mlr.press/v161/eriksson21a/eriksson21a.pdf [Last accessed on 21 Apr 2023]

[65] Wang Q,Huang W,Liu S.Parameterization of NSGA-II for the optimal design of water distribution systems.Water2019;11:971

[66] Hort M.The effect of offspring population size on NSGA-II: a preliminary study. In: Proceedings of the Genetic and Evolutionary Computation Conference Companion; 2021. pp. 179-80.

[67] Tanabe R.The impact of population size, number of children, and number of reference points on the performance of NSGA-III. In: Trautmann H, Rudolph G, Klamroth K, Schütze O, Wiecek M, Jin Y, Grimme C, editors. Evolutionary Multi-Criterion Optimization. Cham: Springer International Publishing; 2017. pp. 606-21.

[68] Jiang Y,Sharma A,Mullin M.An artificial intelligence enabled chemical synthesis robot for exploration and optimization of nanomaterials.Sci Adv2022;8:eabo2626 PMCID:PMC9544322

[69] Liang Q,Ren Z.Benchmarking the performance of Bayesian optimization across multiple experimental materials science domains.NPJ Comput Mater2021;7

[70] Curry DM.Computational complexity measures for many-objective optimization problems.Procedia Comput Sci2014;36:185-91

[71] Kukkonen S.Ranking-dominance and many-objective optimization. In: 2007 IEEE Congress on Evolutionary Computation.IEEE2007:3983-90

[72] Koch P,Emmerich MT,Konen W.Efficient multi-criteria optimization on noisy machine learning problems.App Soft Comput2015;29:357-70

[73] Horn D,Sun X.First Investigations on noisy model-based multi-objective optimization. In: Trautmann H, Rudolph G, Klamroth K, Schütze O, Wiecek M, Jin Y, Grimme C, editors. Evolutionary Multi-Criterion Optimization. Cham: Springer International Publishing; 2017. pp. 298-313.

[74] Kim C,Krishnan S.A hybrid organic-inorganic perovskite dataset.Sci Data2017;4:170057 PMCID:PMC5423391

[75] Häse F,Hickman RJ,Aspuru-guzik A.G ryffin : an algorithm for Bayesian optimization of categorical variables informed by expert knowledge.Appl Phys Rev2021;8:031406