Abstract
A stimulus can be encoded in a population of spiking neurons through any change in the statistics of their spike patterns. Thus, the baseline spike statistics in the absence of a stimulus can impact the population’s encoding capacity. Some neurons maintain a baseline firing pattern and can decrease their spike rate in response to a stimulus. Not only do baseline firing rates vary widely among different types of neurons, but so do their higher-order statistics, like the degree to which they tend to group their spikes together into bursts and how those bursts are grouped across the population. We investigated how higher-order statistics of baseline spike patterns impact how much information a neural population can transmit about a stimulus that drives a gap in firing using a novel information-theoretic decoding mechanism which we call an “information train.” We discover that there is an optimal level of burstiness for gap detection that is robust to other parameters of the neural population like its size, mean firing rate, and level of correlation. We consider this theoretical result in the context of experimental data from different types of retinal ganglion cells with different baseline spike statistics and determine that the spike statistics of bursty suppressed-by-contrast (bSbC) retinal ganglion cells support nearly optimal detection of both the onset and strength of a contrast step.
Competing Interest Statement
The authors have declared no competing interest.
