Optimal quantization in neural coding

Mark D. McDonnell, Derek Abbott

Abstract: In this paper the optimality of the encoding by relaxing the constraint of identical threshold values for each neuron and determining the optimal encoding for a range of SNR's is presented. The population of neurons can be considered a semicontinuous information channel. Using Fisher information that the value of SNR at which bifurcation occurs asymptotically approaches a fixed value of SNR. This result indicates that in the presence of low SNR's, populations of neurons may be able to effectively encode information in a manner similar to a flash analog to digital converter, despite possessing identical thresholds.