In this section we will introduce Learning-Tool by means of an introductory
example. In this simple example we will train a readout neuron modeled
as a threshold gate (see [Hertz et al., 1991]) to classify a spike train. This
readout neuron will receive its input from a neural microcircuit
modeled as a network of leaky-integrate-and-fire neurons (see
[Gerstner and Kistler, 2002]) which is stimulated by the input spike
train (which should be classified). The setup is shown in
Figure1.
Figure:
Architecture used to classify a spike train. The microcircuit
is modeled as a network of leaky-integrate-and-fire neurons.
Two Poisson spike trains (freqency 20Hz, length 0.5sec) are
generated, and fixed as templates 0 and 1. The actual input spike
train is generated as jittered versions of a template by varying each
spike by a random drawn amount (Gaussion distribuion with zero mean
and a given STD; this STD is called jitter (default jitter=4ms)).
The task of the threshold gate is to output the number (0 or 1) of the
(random choosen) template from which the input spike train was
generated.
