A mathematical model for the two-learners problem

Jan Saputra Müller, Carmen Vidaurre, Martijn Schreuder, Frank C. Meinecke, Paul Von Bünau, Klaus Muller

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

Objective. We present the first generic theoretical formulation of the co-adaptive learning problem and give a simple example of two interacting linear learning systems, a human and a machine. Approach. After the description of the training protocol of the two learning systems, we define a simple linear model where the two learning agents are coupled by a joint loss function. The simplicity of the model allows us to find learning rules for both human and machine that permit computing theoretical simulations. Main results. As seen in simulations, an astonishingly rich structure is found for this eco-system of learners. While the co-adaptive learners are shown to easily stall or get out of sync for some parameter settings, we can find a broad sweet spot of parameters where the learning system can converge quickly. It is defined by mid-range learning rates on the side of the learning machine, quite independent of the human in the loop. Despite its simplistic assumptions the theoretical study could be confirmed by a real-world experimental study where human and machine co-adapt to perform cursor control under distortion. Also in this practical setting the mid-range learning rates yield the best performance and behavioral ratings. Significance. The results presented in this mathematical study allow the computation of simple theoretical simulations and performance of real experimental paradigms. Additionally, they are nicely in line with previous results in the BCI literature.

Original languageEnglish
Article number036005
JournalJournal of Neural Engineering
Volume14
Issue number3
DOIs
Publication statusPublished - 2017 Mar 21

Fingerprint

Learning systems
Theoretical Models
Learning
Mathematical models
Linear Models
Joints

Keywords

  • brain-computer interfacing
  • co-adaptation
  • linear models
  • mathematical models
  • theoretical models

ASJC Scopus subject areas

  • Biomedical Engineering
  • Cellular and Molecular Neuroscience

Cite this

Müller, J. S., Vidaurre, C., Schreuder, M., Meinecke, F. C., Von Bünau, P., & Muller, K. (2017). A mathematical model for the two-learners problem. Journal of Neural Engineering, 14(3), [036005]. https://doi.org/10.1088/1741-2552/aa620b

A mathematical model for the two-learners problem. / Müller, Jan Saputra; Vidaurre, Carmen; Schreuder, Martijn; Meinecke, Frank C.; Von Bünau, Paul; Muller, Klaus.

In: Journal of Neural Engineering, Vol. 14, No. 3, 036005, 21.03.2017.

Research output: Contribution to journalArticle

Müller, JS, Vidaurre, C, Schreuder, M, Meinecke, FC, Von Bünau, P & Muller, K 2017, 'A mathematical model for the two-learners problem', Journal of Neural Engineering, vol. 14, no. 3, 036005. https://doi.org/10.1088/1741-2552/aa620b
Müller JS, Vidaurre C, Schreuder M, Meinecke FC, Von Bünau P, Muller K. A mathematical model for the two-learners problem. Journal of Neural Engineering. 2017 Mar 21;14(3). 036005. https://doi.org/10.1088/1741-2552/aa620b
Müller, Jan Saputra ; Vidaurre, Carmen ; Schreuder, Martijn ; Meinecke, Frank C. ; Von Bünau, Paul ; Muller, Klaus. / A mathematical model for the two-learners problem. In: Journal of Neural Engineering. 2017 ; Vol. 14, No. 3.
@article{a8d7320936d64eacaf2c4d0791b9b795,
title = "A mathematical model for the two-learners problem",
abstract = "Objective. We present the first generic theoretical formulation of the co-adaptive learning problem and give a simple example of two interacting linear learning systems, a human and a machine. Approach. After the description of the training protocol of the two learning systems, we define a simple linear model where the two learning agents are coupled by a joint loss function. The simplicity of the model allows us to find learning rules for both human and machine that permit computing theoretical simulations. Main results. As seen in simulations, an astonishingly rich structure is found for this eco-system of learners. While the co-adaptive learners are shown to easily stall or get out of sync for some parameter settings, we can find a broad sweet spot of parameters where the learning system can converge quickly. It is defined by mid-range learning rates on the side of the learning machine, quite independent of the human in the loop. Despite its simplistic assumptions the theoretical study could be confirmed by a real-world experimental study where human and machine co-adapt to perform cursor control under distortion. Also in this practical setting the mid-range learning rates yield the best performance and behavioral ratings. Significance. The results presented in this mathematical study allow the computation of simple theoretical simulations and performance of real experimental paradigms. Additionally, they are nicely in line with previous results in the BCI literature.",
keywords = "brain-computer interfacing, co-adaptation, linear models, mathematical models, theoretical models",
author = "M{\"u}ller, {Jan Saputra} and Carmen Vidaurre and Martijn Schreuder and Meinecke, {Frank C.} and {Von B{\"u}nau}, Paul and Klaus Muller",
year = "2017",
month = "3",
day = "21",
doi = "10.1088/1741-2552/aa620b",
language = "English",
volume = "14",
journal = "Journal of Neural Engineering",
issn = "1741-2560",
publisher = "IOP Publishing Ltd.",
number = "3",

}

TY - JOUR

T1 - A mathematical model for the two-learners problem

AU - Müller, Jan Saputra

AU - Vidaurre, Carmen

AU - Schreuder, Martijn

AU - Meinecke, Frank C.

AU - Von Bünau, Paul

AU - Muller, Klaus

PY - 2017/3/21

Y1 - 2017/3/21

N2 - Objective. We present the first generic theoretical formulation of the co-adaptive learning problem and give a simple example of two interacting linear learning systems, a human and a machine. Approach. After the description of the training protocol of the two learning systems, we define a simple linear model where the two learning agents are coupled by a joint loss function. The simplicity of the model allows us to find learning rules for both human and machine that permit computing theoretical simulations. Main results. As seen in simulations, an astonishingly rich structure is found for this eco-system of learners. While the co-adaptive learners are shown to easily stall or get out of sync for some parameter settings, we can find a broad sweet spot of parameters where the learning system can converge quickly. It is defined by mid-range learning rates on the side of the learning machine, quite independent of the human in the loop. Despite its simplistic assumptions the theoretical study could be confirmed by a real-world experimental study where human and machine co-adapt to perform cursor control under distortion. Also in this practical setting the mid-range learning rates yield the best performance and behavioral ratings. Significance. The results presented in this mathematical study allow the computation of simple theoretical simulations and performance of real experimental paradigms. Additionally, they are nicely in line with previous results in the BCI literature.

AB - Objective. We present the first generic theoretical formulation of the co-adaptive learning problem and give a simple example of two interacting linear learning systems, a human and a machine. Approach. After the description of the training protocol of the two learning systems, we define a simple linear model where the two learning agents are coupled by a joint loss function. The simplicity of the model allows us to find learning rules for both human and machine that permit computing theoretical simulations. Main results. As seen in simulations, an astonishingly rich structure is found for this eco-system of learners. While the co-adaptive learners are shown to easily stall or get out of sync for some parameter settings, we can find a broad sweet spot of parameters where the learning system can converge quickly. It is defined by mid-range learning rates on the side of the learning machine, quite independent of the human in the loop. Despite its simplistic assumptions the theoretical study could be confirmed by a real-world experimental study where human and machine co-adapt to perform cursor control under distortion. Also in this practical setting the mid-range learning rates yield the best performance and behavioral ratings. Significance. The results presented in this mathematical study allow the computation of simple theoretical simulations and performance of real experimental paradigms. Additionally, they are nicely in line with previous results in the BCI literature.

KW - brain-computer interfacing

KW - co-adaptation

KW - linear models

KW - mathematical models

KW - theoretical models

UR - http://www.scopus.com/inward/record.url?scp=85020397428&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85020397428&partnerID=8YFLogxK

U2 - 10.1088/1741-2552/aa620b

DO - 10.1088/1741-2552/aa620b

M3 - Article

C2 - 28224972

AN - SCOPUS:85020397428

VL - 14

JO - Journal of Neural Engineering

JF - Journal of Neural Engineering

SN - 1741-2560

IS - 3

M1 - 036005

ER -