Spiking neurons, receiving temporally encoded inputs, can compute radi
al basis functions (RBFs) by storing the relevant information in their
delays. In this paper we show how these delays can be learned using e
xclusively locally available information (basically the time differenc
e between the pre- and postsynaptic spikes). Our approach gives rise t
o a biologically plausible algorithm for finding clusters in a high-di
mensional input space with networks of spiking neurons, even if the en
vironment is changing dynamically. Furthermore. we show that our learn
ing mechanism makes it possible that such RBF neurons can perform some
kind of feature extraction where they recognize that only certain inp
ut coordinates carry relevant information. Finally we demonstrate that
this model allows the recognition of temporal sequences even if they
are distorted in various ways.