Self-organized Monte Carlo localization of critical point via linear filtering

Self-organized Monte Carlo simulations are suggested. Their essence is artificial dynamics consisting of the well-known single-spin-flip Metropolis algorithm supplemented by biased random walk in temperature space. The action of walker is driven by feedback utilizing the linear filtering recursion based on the instantaneous estimates of Binder cumulants. The simulation for 2d Ising model demonstrates that the mean temperature typical for the steady noncanonical equilibrium regime properly approximates the true critical temperature. The estimates of the critical Binder cumulants and critical exponents are also discussed.