ESR6 - Spiking Neural Networks for Information Representation and Decoding

Using information-theoretic computational analyses of real spike trains, based on the concept of intersection information, we will determine how information about tactile stimuli is carried by neuronal populations, from peripheral to central at various levels, and how the representation at a given stage of processing is read out to give rise to progressively determining the final percept and its use for the task. The metric for the biological validity of this approach (and the devices based upon it) is to use the candidate decoding algorithm in order to specify both the object being contacted by the sensory system and the subjects’ psychophysical choice. Correct decoding of the stimulus indicates that the decoding algorithms have identified information-carrying algorithms, while correct decoding of choice indicates that the algorithms have identified the same elements used by the brain to construct perception. The fellow will therefore develop psychophysical behavioral paradigms for rats and humans in parallel, with methods for fully characterizing motor strategy and sensory input. S/he will record neuronal population data sets at multiple stages of rat tactile processing pathway.

Description of the spike-timing based neural population codes employed for tactile coding and perceptual decisions across the brain. Mathematical extrapolation of these principles to rules to encode information in artificial sensors and to use this information in robots for performing tasks.