Biomedical Engineering Reference
In-Depth Information
a
1
0
0.09
−0.11
b
c
0 200 400 600
Interspike interval (ms)
d
1
0
0.09
−0.11
e
f
0 100 200
Interspike interval (ms)
Figure 12.5
Responses of the nonleaky integrate-and-fire model driven by correlated, binary
noise. The input switches states randomly, but on average the same state is main-
tained for 2t
corr
ms. Sample voltage and input time courses are 50 ms long. Raster
plots show 6 seconds of continuous simulation time. For the three top panels the
correlation time
t
corr
was 1 ms; for the lower panels it was 5 ms. (Adapted from
[69].)
spikes in the model given m and s. The number of spikes separated by this interval
grows as the correlation time increases. At the same time, however, longer correla-
tion times also give rise to long interspike intervals, which occur because the input
can stay in the low state for longer stretches of time. This is why correlation time in-
creases variability: it produces both short and long interspike intervals. The quantity
that is most often used to measure the regularity of a spike train is the coefficient of
variation, or
CV
ISI
, which is equal to the standard deviation of the interspike inter-
vals divided by their mean. The
CV
ISI
in Figures 12.5c is equal to 1, as for a Poisson
process; in Figure 12.5f it is equal to 1.18, which reflects the higher variability. Note
that mand s are the same for all panels. This demonstrates that the input correlation
Search WWH ::
Custom Search