Information Technology Reference
In-Depth Information
Fig. 1.
Topology and spatial connectivity of the LGN model. Arrangement of neurons
in the lattice. Circuit Design of dLGN and number of divergence connections between
different layers.
and
σ
sur
. Receptive field parameters for the ganglion cells have been set in
accordance to Allen and Freeman (2006). The rate of the pulse train evoked by
a stimulus in the ganglion cells is modeled with a 2D convolution function:
r
=
r
0
+
D
(
x, y
)
s
(
x, y
)
dxdy
(3)
where
r
represents the firing rate evoked by a stimulus
s
(
x, y
),
r
0
is the back-
ground firing, and
D
(
x, y
) the receptive field of ganglion cells.
Finally, the firing rate of each ON/OFF ganglion cell modulates a pulse train
generator following the statistics of a homogeneous Poisson process:
P
T
(
n
)=
rT
n
!
e
−rT
(4)
Figure 2 provides an overview of all the components involved in our model.
3.5 Neuron Model
We use an implementation of the Adaptative Exponential Integrate and Fire
(AEIaF) neuron model of Brette and Gerstner [6] as implemented in the NEST
simulator [7]. The AEIaF model is a conductance-based integrate and fire model
with an exponential soft spiking threshold rather than a hard threshold, and with
a second state variable that recreates membrane potential and spike adaptation
effects. With this generic model we can characterize neurons showing similar
Search WWH ::
Custom Search