Canonical transition probabilities for adaptive Metropolis stimulation

We examine non-Boltzmann Monte Carlo algorithms used to study slowly relaxi ng systems. By adding a simple bookkeeping step to the Metropolis algorithm , we obtain statistical estimators of canonical macrostate probabilities. T hese estimators enable a natural accumulation of statistics from simulation s having different importance weights, enable temperature extrapolation wit hout using energy to define macrostate labels, improve parallelization, and reduce variance. We illustrate with an Ising model example.