MLPfit: a simple and powerful tool for training and using Multi-Layer Perceptrons

COMPUTING SEMINAR


Title	:	MLPfit: a simple and powerful tool for training and using Multi-Layer Perceptrons
Speaker	:	Jerome Schwindling / CEA Saclay (FRANCE)
Date	:	Wednesday 19 January 2000 at 16:00hrs
Place	:	IT Auditorium, bldg 31/3-005
Information	:	http://wwwinfo.cern.ch/seminars
Organiser	:	G.Folger / IT

ABSTRACT

Artificial Neural Networks in general, and Multi-Layer Perceptrons in particular, are now commonly used in many commercial, industrial or research fields, including High-Energy physics.

In this seminar, we will first describe some of the mathematical grounds of Multi-Layer Perceptrons (function approximation, minimization methods, linear algebra...). We will then present the MLPfit package, which:

can be used for both function approximation and classification tasks
implements powerful minimization methods in a fast and precise way
is available through various user interfaces, including the latest version of PAW.