An Extension of the Metropolis Algorithm

Abstract The Metropolis algorithm and its variants are perhaps the most widely used methods of generating Markov chains with a specified equilibrium distribution. We study an extension of the Metropolis algorithm from both a decision theoretic and rate of convergence point of view. The decision theoretic approach was first taken by Peskun (1973) who showed some optimality properties of the classical Metropolis sampler. In this article, we propose an extension of the Metropolis algorithm which reduces the asymptotic variance and accelerates the convergence rate of its classic form. The principle method used to improve the properties of a sampler is to move mass from the diagonal elements of the Markov chain's transition matrix to the off-diagonal elements. A low dimensional example is given to illustrate that our extended algorithm converges to stationary distribution in the fastest possible order n steps, while the conventional Metropolis chain takes at least order n 2log(n) steps.

[1]  W. K. Hastings,et al.  Monte Carlo Sampling Methods Using Markov Chains and Their Applications , 1970 .

[2]  N. Popov.  On the Rate of Convergence for Countable Markov Chains , 1980 .

[3]  Charles R. Johnson,et al.  Matrix analysis , 1985, Statistical Inference for Engineers and Data Scientists.

[4]  L. Tierney.  Markov Chains for Exploring Posterior Distributions , 1994 .

[5]  S. Caracciolo,et al.  Nonlocal Monte Carlo algorithm for self-avoiding walks with fixed endpoints , 1990 .

[6]  C. Hwang,et al.  Optimal Spectral Structure of Reversible Stochastic Matrices, Monte Carlo Methods and the Simulation of Markov Random Fields , 1992 .

[7]  G. Casella,et al.  Rao-Blackwellisation of sampling schemes , 1996 .

[8]  P. Diaconis,et al.  Geometric Bounds for Eigenvalues of Markov Chains , 1991 .

[9]  L. Tierney.  A note on Metropolis-Hastings kernels for general state spaces , 1998 .

[10]  Antonietta Mira,et al.  Ordering and Improving the Performance of Monte Carlo Markov Chains , 2001 .

[11]  N. Metropolis,et al.  Equation of State Calculations by Fast Computing Machines , 1953, Resonance.

[12]  Persi Diaconis,et al.  What Do We Know about the Metropolis Algorithm? , 1998, J. Comput. Syst. Sci..

[13]  Jun S. Liu,et al.  Statistical inference and Monte Carlo algorithms , 1996 .

[14]  P. Peskun,et al.  Optimum Monte-Carlo sampling using Markov chains , 1973 .

[15]  P. Diaconis,et al.  COMPARISON THEOREMS FOR REVERSIBLE MARKOV CHAINS , 1993 .

[16]  J. A. Fill.  Eigenvalue bounds on convergence to stationarity for nonreversible markov chains , 1991 .

[17]  Jun S. Liu.  Peskun's theorem and a modified discrete-state Gibbs sampler , 1996 .

[18]  L. Tierney,et al.  Efficiency and Convergence Properties of Slice Samplers , 2002 .