Professor of Image Analysis
School of Electronics Computing and Mathematics
University of Surrey
Guildford, GU2 7XH, United Kingdom
http://www.ee.surrey.ac.uk/Personal/M.Petrou
Title: 3D texture analysis for medical applications
Abstract:
Volume data are very common in Medicine. However, they are difficult to
visualise, yet alone to assess properties like volumetric isotropy or variation.
Currently clinicians see such data slice by slice, they do not and they cannot
take advantage of the full information conveyed by such data. This talk will
present a way of assessing the texture of 3D volume data, a way of visualising
such information, and some applications of the approach to schizophrenia and
Alzheimer's desease.
BellSouth Professor and Director
Computational NeuroEngineering Laboratory
EB 451, Bldg #33
University of Florida
Gainesville, FL 32611
http://www.cnel.ufl.edu/principe/principe.html
Title: Information Theoretic Learning: A Nonparametric Approach
Abstract:
This talk will present a new cost function for adaptation based on Renyi's
entropy. In order to obtain a practical nonparametric cost function for
supervised or unsupervised training of linear or nonlinear mappers, we integrate
Renyi's definition with a Parzen estimator. This estimator of entropy does not
require a data model, and resembles an interaction model for learning (the
information potential).
Properties of the information potential will be presented, and several learning algorithms (batch, stochastic gradient, and a recursive entropy estimator) will be derived. Results in feature extraction, unsupervised clustering, blind source separation and information fitlering will be presented.
Dep. de Electrónica e Telecomunicações
Universidade de Aveiro
3810-193 Aveiro, Portugal
http://www.ieeta.pt/~pjf
Title: Handling impulsive noise
Abstract:
Dealing with impulsive noise remains a challenge, despite the efforts of many
people and the existence of several distinct approaches. This talk addresses the
issue of impulsive noise removal. A few of the possible approaches will be
considered, giving some attention to the two basic frameworks that must be faced
when discussing the problem: the analog case, and the digital case. The
techniques discussed are nonlinear, and under certain conditions lead to the
total removal of the noise. The connections between the techniques discussed and
other approaches (including error control coding) are also discussed.