“…It has been argued that the dynamics and spiking behavior of neurons can be interpreted as an expression of a sampling process by means of which the brain performs such inference operations [7,2]. This approach is appealing because it allows drawing parallels between neural dynamics and wellestablished algorithms such as Markov-Chain Monte-Carlo (MCMC) sampling [5,20,13] or restricted Boltzmann-Machines (RBM) [12,17,22], and various constraint satisfaction problem (CSP) solvers [10,15].…”