Publication Details
Title: Learning From Data: General Issues and Special Applications of Radial Basis Function Networks
Author: A. Baraldi and N. A. Borghese
Group: ICSI Technical Reports
Date: August 1998
PDF: ftp://ftp.icsi.berkeley.edu/pub/techreports/1998/tr-98-028.pdf
Overview:
In the first part of this work some important issues regarding the use of data-driven learning systems are discussed. Next, a special category of learning systems known as artificial Neural Networks (NNs) is presented. Our attention is focused on a specific class of NNs, termed Radial Basis Function (RBF) networks, which are widely employed in classification and function regression tasks. A constructive RBF network, termed Hierarchical RBF (HRBF) model, is proposed. An application where the HRBF model is applied to reconstruct a continuous 3-D surface from range data samples is presented. Keywords: Inductive and deductive types of inference, learning from data, predictive learning, supervised and unsupervised learning, actual risk and empirical risk, curse of dimensionality, basis function, kernel function, neural networks, Multi-Layer-Perceptron, Radial Basis Function network, data-driven and error-driven learning, hybrid learning, one- and two-stage learning, grid-partitioning and scatter-partitioning network, constructive network, Hierarchical Radial Basis Function network.
Bibliographic Information:
ICSI Technical Report TR-98-028
Bibliographic Reference:
A. Baraldi and N. A. Borghese. Learning From Data: General Issues and Special Applications of Radial Basis Function Networks. ICSI Technical Report TR-98-028, August 1998
Author: A. Baraldi and N. A. Borghese
Group: ICSI Technical Reports
Date: August 1998
PDF: ftp://ftp.icsi.berkeley.edu/pub/techreports/1998/tr-98-028.pdf
Overview:
In the first part of this work some important issues regarding the use of data-driven learning systems are discussed. Next, a special category of learning systems known as artificial Neural Networks (NNs) is presented. Our attention is focused on a specific class of NNs, termed Radial Basis Function (RBF) networks, which are widely employed in classification and function regression tasks. A constructive RBF network, termed Hierarchical RBF (HRBF) model, is proposed. An application where the HRBF model is applied to reconstruct a continuous 3-D surface from range data samples is presented. Keywords: Inductive and deductive types of inference, learning from data, predictive learning, supervised and unsupervised learning, actual risk and empirical risk, curse of dimensionality, basis function, kernel function, neural networks, Multi-Layer-Perceptron, Radial Basis Function network, data-driven and error-driven learning, hybrid learning, one- and two-stage learning, grid-partitioning and scatter-partitioning network, constructive network, Hierarchical Radial Basis Function network.
Bibliographic Information:
ICSI Technical Report TR-98-028
Bibliographic Reference:
A. Baraldi and N. A. Borghese. Learning From Data: General Issues and Special Applications of Radial Basis Function Networks. ICSI Technical Report TR-98-028, August 1998