< Back to Article
Orientation Selectivity in Inhibition-Dominated Networks of Spiking Neurons: Effect of Single Neuron Properties and Network Dynamics
Figure 7
Average tuning curves in different networks.
For the network of PIF neurons with 
(A) and 
(B), and for the network of LIF neurons with 
(C), the average tuning curves are shown for comparison. Individual tuning curves are aligned to their Input PO, and the mean and standard deviation of all data points falling in bins of size 
are computed, separately for each contrast. The average tuning curves extracted from simulations are plotted as solid lines (cyan), and the dashed line (magenta) shows the average tuning curve from analytical predictions. The shading indicates the standard deviation, in respective colors. In (B), the average aligned tuning curve from a linear prediction is also plotted for comparison (black dashed lines). Lighter colors correspond to lower contrasts. (D, E, F) For each neuron, the modulation (F2) component of the tuning curve is computed at each contrast. This value for higher contrasts (
) is then scattered against the lowest (
), respectively. The dashed lines denote the result of linear amplification of modulation, 
. The distributions of the respective modulation (F2) ratio 
are plotted in the inset, respectively. (G) In PIF networks with 
, the maximum firing rates of individual tuning curves, 
, are computed. The difference of this maximum value from the simulated results (Diff Max Rate) is then computed (
). The distribution of this difference is plotted for different contrasts, for the linear rectified prediction (NLin. Pred.) and for the linear prediction (Lin. Pred.). Inset: For the middle contrast (
), 
is plotted vs. 
, for both predictions. (H, I) For the PIF network with 
and for different contrasts, the distributions of OSI and dPO (output PO - input PO) are plotted, respectively. For the simulated results and for the linear rectified prediction, the distributions show almost a perfect match. This is compared with the linear prediction (in black).
doi: https://doi.org/10.1371/journal.pcbi.1004045.g007