No label defined (Q4941497)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1413009
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 1413009 |
Statements
12 March 2000
0 references
Statistics
0 references
Neural networks
0 references
neural computing
0 references
applications
0 references
penalized discriminant analysis
0 references
unsupervised learning
0 references
radial basis function networks
0 references
many-parameter models
0 references
density network models
0 references
latent variable models
0 references
data visualisation
0 references
latent structure models
0 references
multivariate statistics
0 references
0.95315266
0 references
0 references
0.92276126
0 references
0.92245036
0 references
0.9224503
0 references
This book covers various contributions from authors who are leading researchers in fields of the interface between statistical research and neural networks. This currently active research interface was the main topic of a workshop in Edinburgh. \textit{T. Hastie}, \textit{Tibshirani} and \textit{A. Buja} create more flexible versions of a relationship between linear discriminant analysis and a sequence of linear regressions by replacing the regressions by splines and multivariate additive regression splines (MARS). This leads to a form of regularised discriminant analysis which is called `penalized discriminant analysis'. \textit{J. Kay} uses information theory for processing in artificial neural networks for unsupervised learning. The networks are composed of local processors which are connected in various architectures. Several versions of the processor are considered. \textit{D. Lowe} reviews some statistical aspects of radial basis function networks with links to conditional and unconditional density estimation. Radial basis function networks show close similarity to kernel based smoothing techniques in statistics. In unsupervised dimension-reducing feature extraction a transformation is given from data space to feature space which is modeled by the radial basis function approach. \textit{N. Intrator} reviews several ways to robustify estimation and prediction in many-parameter models. Concerning variance control models he presents methods for extracting additional information from ensembles of experts by exploiting the variability in responses for estimating the confidence of a certain group of experts. The methods are demonstrated with real-world applications. \textit{D. MacKay} and \textit{M. Gibbs} consider density network models with categorical observables and continuous latent variables as inputs. A probability density is defined which can be trained by various Monte Carlo methods. \textit{Ch.M. Bishop} and \textit{M.E. Tipping} contributed a chapter on latent variable models and data visualisation. \textit{A.P. Dunmur} and \textit{D.M. Titterington} concentrate on latent structure models. Statements of the structures are given, associated with latent profile and latent class models. \(ML\)-estimation using die \(EM\)-algorithm is discussed. Baysian approaches are implemented using versions of the Gibbs sampler. \textit{E.B. Martin} and \textit{A.J. Morris} consider practical problems in industrial processing with the application of various techniques in multivariate statistics and neural networks: Principal components regression, projection to latent structures, feed-forward neural networks, radial basis function neural networks, locally weighted regression.
0 references