NEW RECCURENT NEURAL NETWORK ARCHITECTURES AND NEW TRAINING METHODS FOR FEEDFORWARD NEURAL NETWORKS

Ref.No: 68062700
Start date: 08.11.2002
End date: 31.12.2006
Approval date: 15.09.2004
Department: ELECTRICAL & COMPUTER ENGINEERING
Sector: SIGNALS, CONTROL AND ROBOTICS
Financier: ΕΠΕΑΕΚ II, U.P.E.P.TH
Budget: 31.783,00 €
Scientific Responsible: Prof. NIKOLAOS MARATOS
Email: maratos@ece.ntua.gr
Description: THE PROJECT HAS A TWOFOLD RESEARCH AS FOLLOWS. THE DEVELOPMENT OF NEW RECURRENT NEURAL NETWORK ARCHITECTURES FOR THE SOLUTION OF CONSTRAINED OPTIMIZATION PROBLEMS. THE PROPOSED ARCHITECTURES ARE BASED ON THE GRADIENT PROJECTION AND ON THE RECURSIVE.....
Go to Top