A neural network architecture is introduced which implements a supervi
sed clustering algorithm for the classification of feature vectors. Th
e network is self-organising, and is able to adapt to the shape of the
underlying pattern distribution as well as detect novel input vectors
during training. It is also capable of determining the relative impor
tance of the feature components for classification. The architecture i
s a hybrid of supervised and unsupervised networks, and combines the s
trengths of three well-known architectures: learning vector quantisati
on, back-propagation and adaptive resonance theory. Network performanc
e is compared to that of learning vector quantisation, back-propagatio
n and cascade-correlation. it is found that performance is generally a
s good as or better than the performance of these other architectures,
while training time is considerably shorter. However the main advanta
ge of the hybrid architecture is its ability to gain insight into the
feature pattern space.