Abstract��Neural networks have the power to deal with infor-
mation which is very hard to process using ordinary approaches,
e.g. speech recognition. A recent trend in applying neural net-
works is to use biologically realistic neuron models. Specifically,
neurons are considered which communicate with discrete pulses
instead of continuous signals: spiking neurons. In this paper
we investigate a small selection of properties which are found
in biological neurons and investigate their effect on the general
computational performance of spiking neural networks (SNN).
Firstly, we investigated the way in which the internal dynamics
of the neurons and delayed communication improve the ability
to recognize temporal patterns. Secondly we explored an un-
supervised adaptation rule which helps to distribute the work
equally over all the neurons in the network, so that all neurons
are involved in the task they are supposed to solve. It turned
out that these biologically inspired features often improved the
performance for the tasks investigated.