Collective properties of neural networks: A statistical physics approach (Q2266683)

The collective properties of neural networks are studied by the application of general methods of statistical mechanics. In a situation like neural network, the construction of Hamiltonians is fraught with difficulties. One way is to simulate the equilibrium of large systems; the other consists in modelling the network as a stochastic process evolving in a Markov manner naturally the evolution equation where its general equilibrium solution offers a way of identifying the Hamiltonian. Several models in vogue in physics including the Ising models are reviewed and adapted to the case of assembly of neurons.