Correlation Collapse under Persistent Input Derangement in the Nervous System. Fitting pRAM Models for Thalamo-cortical Circuits

Abstract

We study the functional connections between neurons of thalamo-cortical circuits of the albino rat brain through their spiking time correlation in normal and neuropathic animals. These correlations are learnt by a special artificial neural network, the pRAM hardware, which virtually maps on the physiological neurons’ network, on-line learning conditional probabilities from joint firing records of clusters of neurons. The most relevant probabilities are selected w.r.t. the entropic criterion of efficiently mimicking the actual spiking distribution, and minimal neural maps are drawn on the basis of a strength parameter defined for the synapses as a measure of the spiking interaction between the connected neurons. Growth of the neural activity, together with a shattering of the minimal maps emerge as a distinguishing feature of persistent nociceptive conditions.