Subthreshold information encoding in heterogeneous neural network

In this study, the information encoding issue in the nervous system is investigated. There are lots of factors influencing information encoding in a nervous system. Several of them are neural noise, stimulus intensity, biophysical properties of neurons, types of synaptic connections (electrical or chemical), number of synapses and hence network topology. In recent computational neuroscience studies concerning the information encoding, models designed without considering the heterogeneity, which is a biological reality, have been used. However experimental neuroscience studies performed with electrophysiological and advanced imaging techniques have shown that neurons constituting nervous system are not identical in terms of their electrophysiological and morphological properties. Based on this fact, the information encoding performance of neural populations was investigated in this work, modeling neurons that are the primary elements of the nervous system as heterogeneous elements in the context of excitability. For this purpose, the effects of heterogeneity in the neural excitability on the performance of weak signal detection of neural populations were investigated in the present work. In this study, the relationship between heterogeneity and the signal detection have been studied according to the main features of neuron and stimulus. In first place, mean excitability and then network size have been handled. It is shown that double resonance effect emerges at all values for both features and at two different point heterogeneity can play constructive role in information processing similarly with the effect of noise (Stochastic Resonance) in neural systems. In this study, the effects of the stimulus features such as amplitude and period are also investigated. And lastly, the phenomena of diversity induced resonance is investigated according to properties sinapses. The results obtained from the study show that reckoning heterogeneity of the population provides significant advantages in the processing of neural information.