Cooperative Bayesian Optimization for Imperfect Agents

Ali Khoshvishkaie; Petrus Mikkola; Pierre Alexandre Murena; Samuel Kaski

doi:10.1007/978-3-031-43412-9_28

Cooperative Bayesian Optimization for Imperfect Agents

Ali Khoshvishkaie, Petrus Mikkola, Pierre Alexandre Murena^*, Samuel Kaski

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

5 Citations (Scopus)

Abstract

We introduce a cooperative Bayesian optimization problem for optimizing black-box functions of two variables where two agents choose together at which points to query the function but have only control over one variable each. This setting is inspired by human-AI teamwork, where an AI-assistant helps its human user solve a problem, in this simplest case, collaborative optimization. We formulate the solution as sequential decision-making, where the agent we control models the user as a computationally rational agent with prior knowledge about the function. We show that strategic planning of the queries enables better identification of the global maximum of the function as long as the user avoids excessive exploration. This planning is made possible by using Bayes Adaptive Monte Carlo planning and by endowing the agent with a user model that accounts for conservative belief updates and exploratory sampling of the points to query.

Original language	English
Title of host publication	Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings
Editors	Danai Koutra, Claudia Plant, Manuel Gomez Rodriguez, Elena Baralis, Francesco Bonchi
Publisher	Springer
Pages	475-490
Number of pages	16
ISBN (Print)	978-3-031-43411-2
DOIs	https://doi.org/10.1007/978-3-031-43412-9_28
Publication status	Published - Sept 2023
MoE publication type	A4 Conference publication
Event	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases - Turin, Italy Duration: 18 Sept 2023 → 22 Sept 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher	Springer
Volume	14169 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases
Abbreviated title	ECML PKDD
Country/Territory	Italy
City	Turin
Period	18/09/2023 → 22/09/2023

Funding

This research was supported by EU Horizon 2020 (HumanE AI NET, 952026) and UKRI Turing AI World-Leading Researcher Fellowship (EP/W002973/1). Computational resources were provided by the Aalto Science-IT project from Computer Science IT. The authors would like to thank Prof. Frans Oliehoek and Dr. Mert Celikok for their help in setting up the project and the reviewers for their insightful comments.

Access to Document

10.1007/978-3-031-43412-9_28

https://arxiv.org/abs/2403.04442

Cite this

Khoshvishkaie, A., Mikkola, P., Murena, P. A., & Kaski, S. (2023). Cooperative Bayesian Optimization for Imperfect Agents. In D. Koutra, C. Plant, M. Gomez Rodriguez, E. Baralis, & F. Bonchi (Eds.), Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings (pp. 475-490). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14169 LNAI). Springer. https://doi.org/10.1007/978-3-031-43412-9_28

Khoshvishkaie, Ali ; Mikkola, Petrus ; Murena, Pierre Alexandre et al. / Cooperative Bayesian Optimization for Imperfect Agents. Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings. editor / Danai Koutra ; Claudia Plant ; Manuel Gomez Rodriguez ; Elena Baralis ; Francesco Bonchi. Springer, 2023. pp. 475-490 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{5de2fc0b933d419ca18f91bf2400b7f2,

title = "Cooperative Bayesian Optimization for Imperfect Agents",

abstract = "We introduce a cooperative Bayesian optimization problem for optimizing black-box functions of two variables where two agents choose together at which points to query the function but have only control over one variable each. This setting is inspired by human-AI teamwork, where an AI-assistant helps its human user solve a problem, in this simplest case, collaborative optimization. We formulate the solution as sequential decision-making, where the agent we control models the user as a computationally rational agent with prior knowledge about the function. We show that strategic planning of the queries enables better identification of the global maximum of the function as long as the user avoids excessive exploration. This planning is made possible by using Bayes Adaptive Monte Carlo planning and by endowing the agent with a user model that accounts for conservative belief updates and exploratory sampling of the points to query.",

author = "Ali Khoshvishkaie and Petrus Mikkola and Murena, \{Pierre Alexandre\} and Samuel Kaski",

note = "Funding Information: This research was supported by EU Horizon 2020 (HumanE AI NET, 952026) and UKRI Turing AI World-Leading Researcher Fellowship (EP/W002973/1). Computational resources were provided by the Aalto Science-IT project from Computer Science IT. The authors would like to thank Prof. Frans Oliehoek and Dr. Mert Celikok for their help in setting up the project and the reviewers for their insightful comments. | openaire: EC/H2020/952026/EU//HumanE-AI-Net ; European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD ; Conference date: 18-09-2023 Through 22-09-2023",

year = "2023",

month = sep,

doi = "10.1007/978-3-031-43412-9\_28",

language = "English",

isbn = "978-3-031-43411-2",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "475--490",

editor = "Danai Koutra and Claudia Plant and \{Gomez Rodriguez\}, Manuel and Elena Baralis and Francesco Bonchi",

booktitle = "Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings",

address = "Germany",

}

Khoshvishkaie, A, Mikkola, P, Murena, PA & Kaski, S 2023, Cooperative Bayesian Optimization for Imperfect Agents. in D Koutra, C Plant, M Gomez Rodriguez, E Baralis & F Bonchi (eds), Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14169 LNAI, Springer, pp. 475-490, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, Turin, Italy, 18/09/2023. https://doi.org/10.1007/978-3-031-43412-9_28

Cooperative Bayesian Optimization for Imperfect Agents. / Khoshvishkaie, Ali; Mikkola, Petrus; Murena, Pierre Alexandre et al.
Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings. ed. / Danai Koutra; Claudia Plant; Manuel Gomez Rodriguez; Elena Baralis; Francesco Bonchi. Springer, 2023. p. 475-490 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14169 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Cooperative Bayesian Optimization for Imperfect Agents

AU - Khoshvishkaie, Ali

AU - Mikkola, Petrus

AU - Murena, Pierre Alexandre

AU - Kaski, Samuel

N1 - Funding Information: This research was supported by EU Horizon 2020 (HumanE AI NET, 952026) and UKRI Turing AI World-Leading Researcher Fellowship (EP/W002973/1). Computational resources were provided by the Aalto Science-IT project from Computer Science IT. The authors would like to thank Prof. Frans Oliehoek and Dr. Mert Celikok for their help in setting up the project and the reviewers for their insightful comments. | openaire: EC/H2020/952026/EU//HumanE-AI-Net

PY - 2023/9

Y1 - 2023/9

N2 - We introduce a cooperative Bayesian optimization problem for optimizing black-box functions of two variables where two agents choose together at which points to query the function but have only control over one variable each. This setting is inspired by human-AI teamwork, where an AI-assistant helps its human user solve a problem, in this simplest case, collaborative optimization. We formulate the solution as sequential decision-making, where the agent we control models the user as a computationally rational agent with prior knowledge about the function. We show that strategic planning of the queries enables better identification of the global maximum of the function as long as the user avoids excessive exploration. This planning is made possible by using Bayes Adaptive Monte Carlo planning and by endowing the agent with a user model that accounts for conservative belief updates and exploratory sampling of the points to query.

AB - We introduce a cooperative Bayesian optimization problem for optimizing black-box functions of two variables where two agents choose together at which points to query the function but have only control over one variable each. This setting is inspired by human-AI teamwork, where an AI-assistant helps its human user solve a problem, in this simplest case, collaborative optimization. We formulate the solution as sequential decision-making, where the agent we control models the user as a computationally rational agent with prior knowledge about the function. We show that strategic planning of the queries enables better identification of the global maximum of the function as long as the user avoids excessive exploration. This planning is made possible by using Bayes Adaptive Monte Carlo planning and by endowing the agent with a user model that accounts for conservative belief updates and exploratory sampling of the points to query.

UR - https://www.scopus.com/pages/publications/85174451147

U2 - 10.1007/978-3-031-43412-9_28

DO - 10.1007/978-3-031-43412-9_28

M3 - Conference article in proceedings

AN - SCOPUS:85174451147

SN - 978-3-031-43411-2

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 475

EP - 490

BT - Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings

A2 - Koutra, Danai

A2 - Plant, Claudia

A2 - Gomez Rodriguez, Manuel

A2 - Baralis, Elena

A2 - Bonchi, Francesco

PB - Springer

T2 - European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases

Y2 - 18 September 2023 through 22 September 2023

ER -

Khoshvishkaie A, Mikkola P, Murena PA, Kaski S. Cooperative Bayesian Optimization for Imperfect Agents. In Koutra D, Plant C, Gomez Rodriguez M, Baralis E, Bonchi F, editors, Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings. Springer. 2023. p. 475-490. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-43412-9_28

Cooperative Bayesian Optimization for Imperfect Agents

Abstract

Publication series

Conference

Funding

Access to Document

Other files and links

Fingerprint

-: HumanE-AI-Net, Kaski

Science-IT

Cite this

Cooperative Bayesian Optimization for Imperfect Agents

Abstract

Publication series

Conference

Funding

Access to Document

Other files and links

Fingerprint

Projects

-: HumanE-AI-Net, Kaski

Equipment

Science-IT

Cite this