Adaptive Metropolis-Hastings samplers use information obtained from previous draws to tune the proposal distribution. The tuning is carried out automatically, often repeatedly, and continues after the burn-in period. Because the resulting chain is not Markovian, adaptation needs to be done carefully to ensure convergence to the correct ergodic distribution. In this paper we distill recent theor...

The pairwise influence matrix of Dobrushin has long been used as an analytical tool to bound the rate of convergence of Gibbs sampling. In this work, we use Dobrushin influence as the basis of a practical tool to certify and efficiently improve the quality of a Gibbs sampler. Our Dobrushin-optimized Gibbs samplers (DoGS) offer customized variable selection orders for a given sampling budget and...

This paper addresses the uniform random generation of words from a context-free language (over an alphabet of size k), while constraining every letter to a targeted frequency of occurrence. Our approach consists in an extended – multidimensional – version of the classic Boltzmann samplers [7]. We show that, under mostly strong-connectivity hypotheses, our samplers return a word of size in [(1 −...

Abstract Many analyses on configurable software systems are intractable when confronted with colossal and highly-constrained configuration spaces. These could instead use statistical inference, where a tractable sample accurately predicts results for the entire space. To do so, laws of inference requires each member population to be equally likely included in sample, i.e., sampling process need...

We examine the convergence properties of some simple Gibbs sampler examples under various scans. We find some surprising results, including Gibbs samplers where deterministic-scan is much more efficient than random-scan, and other samplers where the opposite is true. We also present an example where the convergence takes precisely the same time with any fixed deterministic scan, but modifying t...

This paper is devoted to the random generation of particular colored necklaces for which the number of beads of a given color is constrained (these necklaces are called v-balanced). We propose an efficient sampler (its expected time complexity is linear) which satisfies the Boltzmann model principle introduced by Duchon, Flajolet, Louchard and Schaeffer (7). Our main motivation is to show that ...