Each model neuron received approximately 50 synapses,
roughly 1/100 of those exhibited anatomically. Short range inputs arose
from 10% of excitatory neurons within a m
radius and 10% of inhibitory
neurons within a
m radius, thus inhibitory inputs to a column on average
had a broader orientation distribution than excitatory inputs.
Short range synaptic strengths,
,
were 15nS (Ex
Ex), 2.7nS (Ex
In), 35nS
(In
Ex), 0.5 nS (In
In); Ex:
excitatory cell, In: inhibitory cell. Thalamocortical synaptic strengths
were 2.7nS (Th
Ex) and 0.7 nS (Th
In). The lower weight
values reflect fewer synapses anatomically on inhibitory neurons.
Thalamocortical inputs
provided a mild orientation bias to the cells of each column
(gaussian in orientation space with
or a
half-width) which
was sharpened by intracortical connections (cf.
[38]).
The preferred orientation
of the bias to each column (figure 2) was matched to the preferred
cortical orientation from an optically recorded map of
activity in primary visual cortex of the cat
[33].
The total thalamocortical
input to each cell was given by a Poisson process for spike times, where
the mean firing rate was given by the sum of center and surround
contributions.
Each regional contribution was
,
where RF% is the percentage
of the cortical receptive field covered by center or surround.
is the
orientation dependence of the net thalamic input response and
is
the difference between the stimulus and preferred orientations.
Note that thalamic input increases linearly with the log of stimulus
contrast.
Cortical receptive fields (i.e., areas receiving direct thalamic input)
were square and the visual field position
shifted by
for each mm shift of cortical position. The central
stimulus was
by
. The surround stimulus covered
the full field (
by
) except
for the central
by
region.
The central grating orientation was aligned
to maximally activate the central orientation domain (colored green
in figure 2), with a contrast that varied between 1% and 100%.
Surround stimuli were either uniform fields (no surround) or had 100% contrast.
Long range connections were taken to be those longer than m and were
made between columns that differed by less than
in preferred
orientation (see figure 2).
Connection probabilities ranged linearly
between 0.5% for cell pairs with a
difference and 0%
at
difference. 80% of model long range connections were onto excitatory neurons
[28].
Synaptic strengths for long-range excitatory inputs
to excitatory and inhibitory neurons were 1.9nS and 2.3nS, respectively.