Deterministic, Stochastic and Mean-field Annealing of Hopfield Models

  • Deterministic relaxation can get stuck in local minima depending on the starting point.
  • Stochastic relaxation uses simulated annealing at different temperatures with probablistic update, which can help the network get out of the local minima, to settle for a deeper minima.
  • Stochastic relaxation can take time to reach equilibrium.
  • Mean-field annealing is used to speed up the process.