Two Brains Guided Interactive Evolution

A. Kattan, Faiyaz Doctor, M. Arif

Research output: Contribution to conferencePaper

Abstract

In this paper, we show that it is possible to use electroencephalography (EEG) and multi-brain computing with two humans to guide an Interactive Genetic Algorithm (IGA) system. We show that combining neural activity across two brains increases accuracy to guide evolutionary search more effectively. The IGA system involves a simple task of evolving a polygon shape to approximate the shape of a target polygon. Two candidates visually inspected the evolved polygons and mentally ranked them (independently from each other) from 1×10 based on their similarity to the target polygon. In parallel, the IGA system evaluated the fitness of evolved polygons using a standard fitness function. The IGA system was run for a few generations, before evolution was paused and EEG signals were collected from the two candidates. The collected EEG signals were used to train a regression model that received unseen EEG as input and mapped this into fitness values. The trained model was then used to guide the IGA solely by using the EEG signals. Off-line experimental results showed that it was possible to build better regression models that are trained using two EEG signals to capture participants evaluation of fitness. This paper demonstrates the possibility of a new domain of applications for interactive evolution where standard fitness calculations can be replaced with multiple EEG signals for guiding an optimisation process.
Original languageEnglish
Pages3203 - 3208
DOIs
Publication statusPublished - 2015
Event2015 IEEE International Conference on Systems, Man, and Cybernetics - Kowloon, China
Duration: 9 Oct 201512 Oct 2015

Conference

Conference2015 IEEE International Conference on Systems, Man, and Cybernetics
Abbreviated titleSMC
CountryChina
CityKowloon
Period9/10/1512/10/15

Fingerprint

Electroencephalography
Brain
Genetic algorithms

Keywords

  • EEG
  • Interactive Genetic Algorithm
  • multi-brain

Cite this

Kattan, A., Doctor, F., & Arif, M. (2015). Two Brains Guided Interactive Evolution. 3203 - 3208. Paper presented at 2015 IEEE International Conference on Systems, Man, and Cybernetics, Kowloon, China. https://doi.org/10.1109/SMC.2015.556

Two Brains Guided Interactive Evolution. / Kattan, A.; Doctor, Faiyaz; Arif, M.

2015. 3203 - 3208 Paper presented at 2015 IEEE International Conference on Systems, Man, and Cybernetics, Kowloon, China.

Research output: Contribution to conferencePaper

Kattan, A, Doctor, F & Arif, M 2015, 'Two Brains Guided Interactive Evolution' Paper presented at 2015 IEEE International Conference on Systems, Man, and Cybernetics, Kowloon, China, 9/10/15 - 12/10/15, pp. 3203 - 3208. https://doi.org/10.1109/SMC.2015.556
Kattan A, Doctor F, Arif M. Two Brains Guided Interactive Evolution. 2015. Paper presented at 2015 IEEE International Conference on Systems, Man, and Cybernetics, Kowloon, China. https://doi.org/10.1109/SMC.2015.556
Kattan, A. ; Doctor, Faiyaz ; Arif, M. / Two Brains Guided Interactive Evolution. Paper presented at 2015 IEEE International Conference on Systems, Man, and Cybernetics, Kowloon, China.
@conference{2154984101734e789655c358903b32a3,
title = "Two Brains Guided Interactive Evolution",
abstract = "In this paper, we show that it is possible to use electroencephalography (EEG) and multi-brain computing with two humans to guide an Interactive Genetic Algorithm (IGA) system. We show that combining neural activity across two brains increases accuracy to guide evolutionary search more effectively. The IGA system involves a simple task of evolving a polygon shape to approximate the shape of a target polygon. Two candidates visually inspected the evolved polygons and mentally ranked them (independently from each other) from 1×10 based on their similarity to the target polygon. In parallel, the IGA system evaluated the fitness of evolved polygons using a standard fitness function. The IGA system was run for a few generations, before evolution was paused and EEG signals were collected from the two candidates. The collected EEG signals were used to train a regression model that received unseen EEG as input and mapped this into fitness values. The trained model was then used to guide the IGA solely by using the EEG signals. Off-line experimental results showed that it was possible to build better regression models that are trained using two EEG signals to capture participants evaluation of fitness. This paper demonstrates the possibility of a new domain of applications for interactive evolution where standard fitness calculations can be replaced with multiple EEG signals for guiding an optimisation process.",
keywords = "EEG, Interactive Genetic Algorithm, multi-brain",
author = "A. Kattan and Faiyaz Doctor and M. Arif",
year = "2015",
doi = "10.1109/SMC.2015.556",
language = "English",
pages = "3203 -- 3208",
note = "2015 IEEE International Conference on Systems, Man, and Cybernetics, SMC ; Conference date: 09-10-2015 Through 12-10-2015",

}

TY - CONF

T1 - Two Brains Guided Interactive Evolution

AU - Kattan, A.

AU - Doctor, Faiyaz

AU - Arif, M.

PY - 2015

Y1 - 2015

N2 - In this paper, we show that it is possible to use electroencephalography (EEG) and multi-brain computing with two humans to guide an Interactive Genetic Algorithm (IGA) system. We show that combining neural activity across two brains increases accuracy to guide evolutionary search more effectively. The IGA system involves a simple task of evolving a polygon shape to approximate the shape of a target polygon. Two candidates visually inspected the evolved polygons and mentally ranked them (independently from each other) from 1×10 based on their similarity to the target polygon. In parallel, the IGA system evaluated the fitness of evolved polygons using a standard fitness function. The IGA system was run for a few generations, before evolution was paused and EEG signals were collected from the two candidates. The collected EEG signals were used to train a regression model that received unseen EEG as input and mapped this into fitness values. The trained model was then used to guide the IGA solely by using the EEG signals. Off-line experimental results showed that it was possible to build better regression models that are trained using two EEG signals to capture participants evaluation of fitness. This paper demonstrates the possibility of a new domain of applications for interactive evolution where standard fitness calculations can be replaced with multiple EEG signals for guiding an optimisation process.

AB - In this paper, we show that it is possible to use electroencephalography (EEG) and multi-brain computing with two humans to guide an Interactive Genetic Algorithm (IGA) system. We show that combining neural activity across two brains increases accuracy to guide evolutionary search more effectively. The IGA system involves a simple task of evolving a polygon shape to approximate the shape of a target polygon. Two candidates visually inspected the evolved polygons and mentally ranked them (independently from each other) from 1×10 based on their similarity to the target polygon. In parallel, the IGA system evaluated the fitness of evolved polygons using a standard fitness function. The IGA system was run for a few generations, before evolution was paused and EEG signals were collected from the two candidates. The collected EEG signals were used to train a regression model that received unseen EEG as input and mapped this into fitness values. The trained model was then used to guide the IGA solely by using the EEG signals. Off-line experimental results showed that it was possible to build better regression models that are trained using two EEG signals to capture participants evaluation of fitness. This paper demonstrates the possibility of a new domain of applications for interactive evolution where standard fitness calculations can be replaced with multiple EEG signals for guiding an optimisation process.

KW - EEG

KW - Interactive Genetic Algorithm

KW - multi-brain

U2 - 10.1109/SMC.2015.556

DO - 10.1109/SMC.2015.556

M3 - Paper

SP - 3203

EP - 3208

ER -