Keywords: parallel processing, combinatorial optimisation, image processing, neural networks, computer vision, training, HPC
Start Date: 1 February 95 / Duration: 30 months
[ participants / contact]
Although Chile is a fast growing country for R&D, parallel processing technologies have not spread yet. However, parallel processing is getting more and more important for industrial applications, especially in terms of cost/performance. There is an important need for research, training and education courses in this field.
The main objective of the project is the effective use of high performance computers in the resolution of real problems for the mining and forestry industry, two of the principal economic activities in the country. Also, special training and research collaboration actions with the European partners will be developed. The research will be essentially focused on physical problems modelling, image processing, optimisation algorithms, neural networks, as well as their applications to some industrial and theoretical problems.
The installation of the MATRA-CAPITAN machine took place in April 1996. On going work includes the development of a parallel simulation of some elementary particles systems, the development of a parallel back propagation algorithm for neural networks and the development of an elementary genetic algorithm for the architecture installed.
Info material concerning the machine installed and its capabilities as well as documents for the Chilean custom services have been prepared. An article concerning the use of this computer for simulations of production methods has been published in a local newspaper.
Universidad de Chile
Facultad de Ingenieria
Matematica
Beaucheff 860, Casilla 170-3, Santiago, CL
EU Partners
Matra Cap Systems, F
LIP-ENS Lyon, F
Polytechnical University
of Ctalaunya, E
Non-EU Partners
Universidad de Chile, CL
CIMM, CL
Arauco, CL
Dr. Eric Goles
Tel: +56 2 671 76 01
Fax: +56 2 209 67
29
E-mail: egoles@uchcecvm.cec.uchile.cl
PARANDES - ITDC-147, May 1997
please address enquiries to the ESPRIT Information Desk
html version of synopsis by Nick Cook