Abstract:

Classic studies show that in many species - from leech and cricket to primate - responses of neural populations can be quite successfully read out using a measure neural population activity termed the population vector. However, despite its successes, detailed analyses have shown that the standard population vector discards substantial amounts of information contained in the responses of a neural population, and so is unlikely to accurately describe how signal communication between parts of the nervous system. I will describe recent theoretical results showing how to modify the population vector expression in order to read out neural responses without information loss. Compared to the standard population vector, the information-preserving read out includes just one additional weighting factor that describes the sharpness of neuronal nonlinearity and represents a measure of neuronal variability. In this way, more reliable neurons are weighted more than weakly tuned neurons. It is noteworthy that there is no simple expression for the information-preserving read-out when written in terms of parameters of neural tuning curves. Although noise correlation affect the amount of the information contained in the responses of the neural population, the same read-out expression continues to work when noise correlations increase or decrease in strength. These results demonstrate how to quantify information transmitted by neurons with irregular tuning curves. I will describe three approximations that make it possible to quantify information transmitted by large neural populations containing thousands of neurons.