Abstract
To crack the neural code and read out the information neural spikes convey, it is essential to understand how the information is coded and how much of it is available for decoding. To this end, it is indispensable to derive from first principles a minimal set of spike features containing the complete information content of a neuron. Here we present such a complete set of coding features. We show that temporal pairwise spike correlations fully determine the information conveyed by a single spiking neuron with finite temporal memory and stationary spike statistics. We reveal that interspike interval temporal correlations, which are often neglected, can significantly change the total information. Our findings provide a conceptual link between numerous disparate observations and recommend shifting the focus of future studies from addressing firing rates to addressing pairwise spike correlation functions as the primary determinants of neural information.
Highlights
To crack the neural code and read out the information neural spikes convey, it is essential to understand how the information is coded and how much of it is available for decoding
Addressing first the case of the Poissonian analogue, we find that the stimulus induced rate modulation is captured by the cross-correlation function, which appears in the numerator of equation (4) and which is equal to the autocorrelation function of the peristimulus time histogram (PSTH)
The list of spike timing features that have been implicated in neural coding includes the average number of spikes per time[5] or the occurrence frequency of spike doublets or triplets[6]. This list has experienced unprecedented growth in the last years as interactions between two, three or even N time points have increasingly been linked to neural information content[17,33,34,35]
Summary
To crack the neural code and read out the information neural spikes convey, it is essential to understand how the information is coded and how much of it is available for decoding To this end, it is indispensable to derive from first principles a minimal set of spike features containing the complete information content of a neuron. At the same time, growing experimental evidence offers a potential solution and indicates that neural activity does not explore all possible combinations but unfolds in many different brain regions along task-specific low-dimensional subregions[4] To illustrate this typical coding situation, we sketch schematically a low-dimensional information coding sub-space on the background of the full phase space of all possible spiking combinations. Motivated by the highly variable neural activity across time and repetitions[1], and the observation of irregular but stable firing across time in many experiments[7,10,11], we focused on the information coding in irregular, stationary spike trains with finite memory, which are supported by experimental in vivo evidence and are the cornerstone of current theoretical approaches[3,5,7,12]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
