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

^*Tämän työn vastaava kirjoittaja

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussa › Conference article in proceedings › Scientific › vertaisarvioitu

2 Sitaatiot (Scopus)

Abstrakti

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.

Alkuperäiskieli	Englanti
Otsikko	Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings
Toimittajat	Danai Koutra, Claudia Plant, Manuel Gomez Rodriguez, Elena Baralis, Francesco Bonchi
Kustantaja	Springer
Sivut	475-490
Sivumäärä	16
ISBN (painettu)	978-3-031-43411-2
DOI - pysyväislinkit	https://doi.org/10.1007/978-3-031-43412-9_28
Tila	Julkaistu - syysk. 2023
OKM-julkaisutyyppi	A4 Artikkeli konferenssijulkaisussa
Tapahtuma	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases - Turin, Italia Kesto: 18 syysk. 2023 → 22 syysk. 2023

Julkaisusarja

Nimi	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Kustantaja	Springer
Vuosikerta	14169 LNAI
ISSN (painettu)	0302-9743
ISSN (elektroninen)	1611-3349

Conference

Conference	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases
Lyhennettä	ECML PKDD
Maa/Alue	Italia
Kaupunki	Turin
Ajanjakso	18/09/2023 → 22/09/2023

Pääsy asiakirjaan

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

https://arxiv.org/abs/2403.04442

Muut tiedostot ja linkit

Linkki julkaisuun Scopuksessa

-: HumanE-AI-Net, Kaski
Kaski, S. (Vastuullinen tutkija), Mikkola, P. (Projektin jäsen), Khoshvishkaie, A. (Projektin jäsen), Hämäläinen, A. (Projektin jäsen), Ibrahim, N. (Projektin jäsen) & Zhu, Y. (Projektin jäsen)
01/09/2020 → 31/08/2024
Projekti: EU: Framework programmes funding

Science-IT
Hakala, M. (Manager)
Perustieteiden korkeakoulu
Laitteistot/tilat: Facility

Siteeraa tätä

Khoshvishkaie, A., Mikkola, P., Murena, P. A., & Kaski, S. (2023). Cooperative Bayesian Optimization for Imperfect Agents. teoksessa D. Koutra, C. Plant, M. Gomez Rodriguez, E. Baralis, & F. Bonchi (Toimittajat), Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings (Sivut 475-490). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vuosikerta 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. Toimittaja / Danai Koutra ; Claudia Plant ; Manuel Gomez Rodriguez ; Elena Baralis ; Francesco Bonchi. Springer, 2023. Sivut 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. julkaisussa D Koutra, C Plant, M Gomez Rodriguez, E Baralis & F Bonchi (toim), 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), Vuosikerta 14169 LNAI, Springer, Sivut 475-490, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, Turin, Italia, 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. toim. / Danai Koutra; Claudia Plant; Manuel Gomez Rodriguez; Elena Baralis; Francesco Bonchi. Springer, 2023. s. 475-490 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vuosikerta 14169 LNAI).

Tutkimustuotos: Artikkeli kirjassa/konferenssijulkaisussa › Conference article in proceedings › Scientific › vertaisarvioitu

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 - http://www.scopus.com/inward/record.url?scp=85174451147&partnerID=8YFLogxK

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. julkaisussa Koutra D, Plant C, Gomez Rodriguez M, Baralis E, Bonchi F, toimittajat, Machine Learning and Knowledge Discovery in Databases : Research Track - European Conference, ECML PKDD 2023, Proceedings. Springer. 2023. s. 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

Abstrakti

Julkaisusarja

Conference

Pääsy asiakirjaan

Muut tiedostot ja linkit

Sormenjälki

Projektit

-: HumanE-AI-Net, Kaski

Laitteet

Science-IT

Siteeraa tätä