The onset latency of single neuron responses in the visual system depends strongly on stimulus contrast. While in V1 latency increases are tens of ms, in higher visual areas (IT) the latency can increase 200 ms at the lowest contrast (Oram et al, 2002 Philosophical Transactions of the Royal Society B: Biological Sciences 357 987 - 1001). We present a layered neural network model of noisy integrate-and-fire neurons. Crucially, the model has strong recurrent connectivity and synapses with short-term synaptic depression. With these realistic ingredients, the model reproduces the contrast-dependent latencies. The model furthermore predicts a strong dependence of the spiking statistics on the contrast and time after stimulus onset. We analysed the response statistics predicted by the model and compared them to the data. The study shows that recurrence and short-term synaptic depression are important to explain dynamics and statistics of visually evoked responses in higher visual areas.
|Publication status||Published - 2007|
|Event||30th European Conference on Visual Perception - Arezzo, Italy|
Duration: 27 Aug 2007 → 31 Aug 2007
|Conference||30th European Conference on Visual Perception|
|Period||27/08/07 → 31/08/07|