Heterogeneity in normalization and attentional modulation in a circuit model

The size of a neuron’s receptive field increases along the visual hierarchy. Neurons in higher-order visual areas integrate information through a canonical mechanism called normalization, where neurons respond sublinearly to multiple stimuli in the receptive field (Carandini & Heeger, 2012). Neurons in visual cortex exhibit highly heterogeneous degrees of normalization, which can be measured by normalization index, defined as the sum of a neuron’s firing rate to each one of two stimuli divided by its firing rate when both stimuli are presented. Interestingly, it has been demonstrated that the heterogeneity in normalization is also strongly associated with the heterogeneity in attentional modulation in firing rates (Ni et al., 2012). However, the circuit mechanism underlying such heterogeneity is still unclear. In this work, we adopt a two-layer spiking neuron circuit model, modeling V1 and V4 areas, to capture the effects of normalization and attentional modulation on both trial-average and trial-variable responses of a neural population. Our circuit model naturally produces a wide distribution of normalization indexes across neurons. The normalization index of each neuron is highly correlated with the magnitude of the inhibitory current it receives. In addition, we find that the pairwise correlation between two neurons is related to their normalization indexes. Specifically, neurons with similar normalization indexes have higher correlation than those with different indexes, consistent with a recent experimental finding (Ruff et al., 2016). Further, our modeled attention improves the communication from the attended V1 location to V4 neurons. In particular, V4 neurons with larger normalization indexes are better predicted by V1 neurons from the attended location. Together, our model captures several experimental findings on normalization and attentional modulation in visual cortex, and we identify inhibition to be the main determinant of the heterogeneity in these modulations.