RANDOM FIELD AND NEURAL INFORMATION

A representation - the Ear-Th representation - of the activity of an assembly of neurons is proposed that allows us to describe simultaneous recorded spike trains through the concept of the random field. This representation intrinsically takes into account the fundamental properties of the neuronal signal: its temporal, stochastic, and spatial nature. As a consequence, a neural network, considered as a kind of parallel random automata, delivers an output random field in response to the excitation provided by a random field that represents the activity of some input fibers. Each random field is represented by its associated Gibbs measure, whose potential is plotted in our representation. This approach is applied to the modeling of an intermediary neural network, which receives its input excitation from the auditory nerve fibers and delivers its response to the next auditory neuronal layer.