* Corresponding author. E-mail address: moises@ugr.es (M. Salmero H n).  Partially supported by CICYT, Spain, under Project TIC97-1149.  Partially supported by CICYT, Spain, under Project TIC98-0982. Neurocomputing 41 (2001) 153}172 Improved RAN sequential prediction using orthogonal techniques Moise H s Salmero H n*, Julio Ortega, Carlos G. Puntonet, Alberto Prieto Departamento de Arquitectura y Tecnologn & a de Computadores, Universidad de Granada, 18071, E-18071 Granada, Spain Accepted 4 December 2000 Abstract A new learning strategy for time-series prediction using radial basis function (RBF) networks is introduced. Its potential is examined in the particular case of the resource allocating network model, although the same ideas could apply to extend any other procedure. In the early stages of learning, addition of successive new groups of RBFs provides an increased rate of conver- gence. At the same time, the optimum lag structure is determined using orthogonal techniques such as QR factorization and singular value decomposition (SVD). We claim that the same techniques can be applied to the pruning problem, and thus they are a useful tool for compaction of information. Our comparison with the original RAN algorithm shows a compa- rable error measure but much smaller-sized networks. The extra e!ort required by QR and SVD is balanced by the simplicity of only using least mean squares for the iterative parameter adaptation.  2001 Elsevier Science B.V. All rights reserved. Keywords: QR decomposition; Radial basis function networks; Resource allocating network; Singular value decomposition; Time-series prediction 1. Introduction In this paper we consider a solution to prediction problems using a learning scheme for arti"cial neural networks known as resource allocating network (RAN) [11]. 0925-2312/01/$-see front matter  2001 Elsevier Science B.V. All rights reserved. PII:S0925-2312(00)00363-5