Thermodynamics for a network of neurons: Signatures of criticality

摘要: The activity of a neural network is defined by patterns spiking and silence from the individual neurons. Because spikes are (relatively) sparse, with increasing numbers less probable, but more number possible increases. This tradeoff between probability numerosity mathematically equivalent to relationship entropy energy in statistical physics. We construct this for populations up N=160 neurons small patch vertebrate retina, using combination direct model-based analyses experiments on response naturalistic movies. see signs thermodynamic limit, where per neuron approaches smooth function as N form corresponds distribution being poised near an unusual kind critical point. Networks or correlation among would not reach state. suggest further tests criticality, give brief discussion its functional significance.