| One of the most fundamental functions of brains is to process information. Whether we are engaged in tasks like reading a book, listening to our favorite music station on radio, smelling a flower in bloom or relishing our favorite gourmet cuisine, we invariably employ our brains to process the information received through our senses and create a perception of the world around us. The physical signals incident on our sensory organs, either in the form of photon fluxes, acoustic vibrations, or plumes of chemical concentrations, are transduced, represented and processed as electrical signals within our brains. One of the essential inquiries in neuroscience is the nature of this representation of information in the brain. This is often referred to as the "neural coding" problem which has been and continues to be the object of a lot of theoretical and experimental scientific effort.; In most theoretical approaches that address the problem, nerve cells are characterized empirically by collection of their input-output responses. The knowledge of constraints imposed on information processing due to biophysics of the underlying biological hardware is generally ignored. This thesis reports the outcome of our efforts to combine techniques from stochastic processes, information theory and single neuron biophysics to unravel the neural coding problem. We believe that a systematic reductionist analysis which takes into account the extant noise due to biological processes specific to neuronal processing will provide fundamental insights overlooked in earlier approaches. We analytically characterize the sources of biological noise associated with different stages in the neuronal information pathway, namely the synapse, the dendritic tree and the spike-initiation zone and employ information-theoretical measures to compute the ability of these components to transmit information in specific signal processing tasks. For analytical tractability, we demonstrate our results using abstract and simplified mathematical models. However, our approach can be readily applied to realistic and complicated descriptions of single neurons to provide a greater understanding of the role of noise in neuronal communication. |
