Friday, May 14

Approximation by Ridge Functions: Neural Networks

4:15 PM-6:15 PM
Room: Savannah 2

Single hidden-layer feedforward neural networks are widely applied to approximation or prediction in applied sciences. In this approach, one approximates a multivariate target function by a sum of ridge functions; this is similar to projection pursuit in the literature of statistics. This poses new and challenging questions both at a practical and theoretical level, ranging from the construction of neural networks to their efficiency and capability. The speakers in this minisymposium will present recent achievements in this area from both theoretical and practical viewpoints.

Organizers: David L. Donoho and Emmanuel J. Candes
Stanford University

4:15-4:40 Ridge Function Approximation: History, Motivation, Overview
David L. Donoho, Organizer
4:45-5:10 Ridgelets: Theory and Applications
Emmanuel J. Candes, Organizer
5:15-5:40 Approximation by Ridge Functions and Neural Networks
Pencho Petrushev, University of South Carolina, Columbia
5:45-6:10 On Local Greedy Approximation and Estimation by Ridge Functions
Lee K. Jones, University of Massachussets, Lowell

AN99 Home


Program Updates

Speaker Index




tjf, 1/19/99, MMD, 2/2/99