Error-driven active learning in growing radial basis function networks for early robot learning,

Qinggang Meng; M. H. Lee

doi:10.1016/j.neucom.2007.05.012

Error-driven active learning in growing radial basis function networks for early robot learning,

Standard

Error-driven active learning in growing radial basis function networks for early robot learning, / Meng, Qinggang; Lee, M. H.

In: Neurocomputing, Vol. 71, No. 7-9, 03.2008, p. 1449-1461.

Research output: Contribution to journal › Article › peer-review

Bibtex - Download

@article{768da8aa1123482680333838f2ca017c,

title = "Error-driven active learning in growing radial basis function networks for early robot learning,",

abstract = "In this paper, we describe a new error-driven active learning approach to self-growing radial basis function networks for early robot learning. There are several mappings that need to be set up for an autonomous robot system for sensorimotor coordination and transformation of sensory information from one modality to another, and these mappings are usually highly nonlinear. Traditional passive learning approaches usually cause both large mapping errors and nonuniform mapping error distribution compared to active learning. A hierarchical clustering technique is introduced to group large mapping errors and these error clusters drive the system to actively explore details of these clusters. Higher level local growing radial basis function subnetworks are used to approximate the residual errors from previous mapping levels. Plastic radial basis function networks construct the substrate of the learning system and a simplified node-decoupled extended Kalman filter algorithm is presented to train these radial basis function networks. Experimental results are given to compare the performance among active learning with hierarchical adaptive RBF networks, passive learning with adaptive RBF networks and hierarchical mixtures of experts, as well as their robustness under noise conditions.",

keywords = "Biologically inspired robotics, Active learning, Hierarchical adaptive radial basis function networks",

author = "Qinggang Meng and Lee, {M. H.}",

note = "Progress in Modeling, Theory, and Application of Computational Intelligence — 15th European Symposium on Artificial Neural Networks 2007 15th European Symposium on Artificial Neural Networks 2007 ",

year = "2008",

month = mar,

doi = "10.1016/j.neucom.2007.05.012",

language = "English",

volume = "71",

pages = "1449--1461",

journal = "Neurocomputing",

issn = "0925-2312",

publisher = "Elsevier",

number = "7-9",

}

RIS (suitable for import to EndNote) - Download

TY - JOUR

T1 - Error-driven active learning in growing radial basis function networks for early robot learning,

AU - Meng, Qinggang

AU - Lee, M. H.

N1 - Progress in Modeling, Theory, and Application of Computational Intelligence — 15th European Symposium on Artificial Neural Networks 2007 15th European Symposium on Artificial Neural Networks 2007

PY - 2008/3

Y1 - 2008/3

N2 - In this paper, we describe a new error-driven active learning approach to self-growing radial basis function networks for early robot learning. There are several mappings that need to be set up for an autonomous robot system for sensorimotor coordination and transformation of sensory information from one modality to another, and these mappings are usually highly nonlinear. Traditional passive learning approaches usually cause both large mapping errors and nonuniform mapping error distribution compared to active learning. A hierarchical clustering technique is introduced to group large mapping errors and these error clusters drive the system to actively explore details of these clusters. Higher level local growing radial basis function subnetworks are used to approximate the residual errors from previous mapping levels. Plastic radial basis function networks construct the substrate of the learning system and a simplified node-decoupled extended Kalman filter algorithm is presented to train these radial basis function networks. Experimental results are given to compare the performance among active learning with hierarchical adaptive RBF networks, passive learning with adaptive RBF networks and hierarchical mixtures of experts, as well as their robustness under noise conditions.

AB - In this paper, we describe a new error-driven active learning approach to self-growing radial basis function networks for early robot learning. There are several mappings that need to be set up for an autonomous robot system for sensorimotor coordination and transformation of sensory information from one modality to another, and these mappings are usually highly nonlinear. Traditional passive learning approaches usually cause both large mapping errors and nonuniform mapping error distribution compared to active learning. A hierarchical clustering technique is introduced to group large mapping errors and these error clusters drive the system to actively explore details of these clusters. Higher level local growing radial basis function subnetworks are used to approximate the residual errors from previous mapping levels. Plastic radial basis function networks construct the substrate of the learning system and a simplified node-decoupled extended Kalman filter algorithm is presented to train these radial basis function networks. Experimental results are given to compare the performance among active learning with hierarchical adaptive RBF networks, passive learning with adaptive RBF networks and hierarchical mixtures of experts, as well as their robustness under noise conditions.

KW - Biologically inspired robotics

KW - Active learning

KW - Hierarchical adaptive radial basis function networks

UR - http://hdl.handle.net/2160/8352

U2 - 10.1016/j.neucom.2007.05.012

DO - 10.1016/j.neucom.2007.05.012

M3 - Article

VL - 71

SP - 1449

EP - 1461

JO - Neurocomputing

JF - Neurocomputing