Learning regions for building a world model from clusters in probability distributions

Witold Slowinski, Frank Guerin

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

Abstract—A developing agent learns a model of the world
by observing regularities occurring in its sensory inputs. In a
continuous domain where the model is represented by a set of
rules, a significant part of the task of learning such a model is to
find appropriate intervals within the continuous state variables,
such that these intervals can be used to define rules whose
predictions are reliable. We propose a technique to find such
intervals (or regions) by means of finding clusters on approximate
probability distributions of sensory variables. We compare
this cluster-based method with an alternative landmark-based
algorithm. We evaluate both techniques on a data log recorded
in a simulation based on OpenArena, a three-dimensional firstperson-
perspective computer game, and demonstrate the results
of how the techniques can learn rules which describe walking
behaviour. While both techniques work reasonably well, the
clustering approach seems to give more “natural” regions which
correspond more closely to what a human would expect; we
speculate that such regions should be more useful if they are to
form a basis for further learning of higher order rules.
Original languageEnglish
Title of host publication2011 IEEE International Conference on Development and Learning (ICDL 2011)
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
Pagesn/a
Number of pages8
Volume2
ISBN (Print)978-1-61284-989-8
DOIs
Publication statusPublished - 24 Aug 2011
EventInternational Conference on Development and Learning and Epigenetic Robotics - Frankfurt, Germany
Duration: 24 Aug 2011 → …

Conference

ConferenceInternational Conference on Development and Learning and Epigenetic Robotics
CountryGermany
CityFrankfurt
Period24/08/11 → …

Fingerprint

Probability distributions
Computer games

Cite this

Slowinski, W., & Guerin, F. (2011). Learning regions for building a world model from clusters in probability distributions. In 2011 IEEE International Conference on Development and Learning (ICDL 2011) (Vol. 2, pp. n/a). Institute of Electrical and Electronics Engineers (IEEE). https://doi.org/10.1109/DEVLRN.2011.6037339

Learning regions for building a world model from clusters in probability distributions. / Slowinski, Witold; Guerin, Frank.

2011 IEEE International Conference on Development and Learning (ICDL 2011). Vol. 2 Institute of Electrical and Electronics Engineers (IEEE), 2011. p. n/a.

Research output: Chapter in Book/Report/Conference proceedingChapter

Slowinski, W & Guerin, F 2011, Learning regions for building a world model from clusters in probability distributions. in 2011 IEEE International Conference on Development and Learning (ICDL 2011). vol. 2, Institute of Electrical and Electronics Engineers (IEEE), pp. n/a, International Conference on Development and Learning and Epigenetic Robotics, Frankfurt, Germany, 24/08/11. https://doi.org/10.1109/DEVLRN.2011.6037339
Slowinski W, Guerin F. Learning regions for building a world model from clusters in probability distributions. In 2011 IEEE International Conference on Development and Learning (ICDL 2011). Vol. 2. Institute of Electrical and Electronics Engineers (IEEE). 2011. p. n/a https://doi.org/10.1109/DEVLRN.2011.6037339
Slowinski, Witold ; Guerin, Frank. / Learning regions for building a world model from clusters in probability distributions. 2011 IEEE International Conference on Development and Learning (ICDL 2011). Vol. 2 Institute of Electrical and Electronics Engineers (IEEE), 2011. pp. n/a
@inbook{c10c6c98e7bd42ea824d5856cdc0a4b7,
title = "Learning regions for building a world model from clusters in probability distributions",
abstract = "Abstract—A developing agent learns a model of the world by observing regularities occurring in its sensory inputs. In a continuous domain where the model is represented by a set of rules, a significant part of the task of learning such a model is to find appropriate intervals within the continuous state variables, such that these intervals can be used to define rules whose predictions are reliable. We propose a technique to find such intervals (or regions) by means of finding clusters on approximate probability distributions of sensory variables. We compare this cluster-based method with an alternative landmark-based algorithm. We evaluate both techniques on a data log recorded in a simulation based on OpenArena, a three-dimensional firstperson- perspective computer game, and demonstrate the results of how the techniques can learn rules which describe walking behaviour. While both techniques work reasonably well, the clustering approach seems to give more “natural” regions which correspond more closely to what a human would expect; we speculate that such regions should be more useful if they are to form a basis for further learning of higher order rules.",
author = "Witold Slowinski and Frank Guerin",
year = "2011",
month = "8",
day = "24",
doi = "10.1109/DEVLRN.2011.6037339",
language = "English",
isbn = "978-1-61284-989-8",
volume = "2",
pages = "n/a",
booktitle = "2011 IEEE International Conference on Development and Learning (ICDL 2011)",
publisher = "Institute of Electrical and Electronics Engineers (IEEE)",

}

TY - CHAP

T1 - Learning regions for building a world model from clusters in probability distributions

AU - Slowinski, Witold

AU - Guerin, Frank

PY - 2011/8/24

Y1 - 2011/8/24

N2 - Abstract—A developing agent learns a model of the world by observing regularities occurring in its sensory inputs. In a continuous domain where the model is represented by a set of rules, a significant part of the task of learning such a model is to find appropriate intervals within the continuous state variables, such that these intervals can be used to define rules whose predictions are reliable. We propose a technique to find such intervals (or regions) by means of finding clusters on approximate probability distributions of sensory variables. We compare this cluster-based method with an alternative landmark-based algorithm. We evaluate both techniques on a data log recorded in a simulation based on OpenArena, a three-dimensional firstperson- perspective computer game, and demonstrate the results of how the techniques can learn rules which describe walking behaviour. While both techniques work reasonably well, the clustering approach seems to give more “natural” regions which correspond more closely to what a human would expect; we speculate that such regions should be more useful if they are to form a basis for further learning of higher order rules.

AB - Abstract—A developing agent learns a model of the world by observing regularities occurring in its sensory inputs. In a continuous domain where the model is represented by a set of rules, a significant part of the task of learning such a model is to find appropriate intervals within the continuous state variables, such that these intervals can be used to define rules whose predictions are reliable. We propose a technique to find such intervals (or regions) by means of finding clusters on approximate probability distributions of sensory variables. We compare this cluster-based method with an alternative landmark-based algorithm. We evaluate both techniques on a data log recorded in a simulation based on OpenArena, a three-dimensional firstperson- perspective computer game, and demonstrate the results of how the techniques can learn rules which describe walking behaviour. While both techniques work reasonably well, the clustering approach seems to give more “natural” regions which correspond more closely to what a human would expect; we speculate that such regions should be more useful if they are to form a basis for further learning of higher order rules.

U2 - 10.1109/DEVLRN.2011.6037339

DO - 10.1109/DEVLRN.2011.6037339

M3 - Chapter

SN - 978-1-61284-989-8

VL - 2

SP - n/a

BT - 2011 IEEE International Conference on Development and Learning (ICDL 2011)

PB - Institute of Electrical and Electronics Engineers (IEEE)

ER -