Enhancing sampling in computational statistics using modified hamiltonians

The Hamiltonian Monte Carlo (HMC) method has been recognized as a powerful sampling tool in computational statistics. In this thesis,we showthat performance ofHMCcan be dramatically improved by replacing Hamiltonians in theMetropolis test with modified Hamiltonians, and a complete momentum update with a partial momentum refreshment. The resulting generalized HMC importance sampler, whichwe called Mix & Match Hamiltonian Monte Carlo (MMHMC), arose as an extension of the Generalized Shadow Hybrid Monte Carlo (GSHMC) method, previously proposed for molecular simulation. The MMHMC method adapts GSHMC specifically to computational statistics and enriches it with new essential features: (i) the e icient algorithms for computation of modified Hamiltonians; (ii) the implicit momentum update procedure and (iii) the two-stage splitting integration schemes specially derived for the methods sampling with modified Hamiltonians. In addition, di erent optional strategies formomentumupdate and flipping are introduced as well as algorithms for adaptive tuning of parameters and e icient sampling of multimodal distributions are developed. MMHMChas been implemented in the in-house so ware package HaiCS (Hamiltonians in Computational Statistics) written in C, tested on the popular statistical models and compared in sampling e iciency with HMC, Generalized Hybrid Monte Carlo, Riemann Manifold Hamiltonian Monte Carlo, Metropolis Adjusted Langevin Algorithm and RandomWalk Metropolis-Hastings. The analysis of time-normalized e ective sample size reveals the superiority of MMHMC over popular sampling techniques, especially in solving high-dimensional problems.

[1]  E. Hairer,et al.  Geometric Numerical Integration , 2022, Oberwolfach Reports.

[2]  John Salvatier,et al.  Probabilistic programming in Python using PyMC3 , 2016, PeerJ Comput. Sci..

[3]  L. Carin,et al.  Monomial Gamma Monte Carlo Sampling , 2016 .

[4]  Adrian Sandu,et al.  A Hybrid Monte Carlo Sampling Filter for Non-Gaussian Data Assimilation , 2015 .

[5]  J. M. Sanz-Serna,et al.  Randomized Hamiltonian Monte Carlo , 2015, 1511.09382.

[6]  D. Dunson,et al.  Recycling Intermediate Steps to Improve Hamiltonian Monte Carlo , 2015, Bayesian Analysis.

[7]  Ingmar Schuster,et al.  Gradient Importance Sampling , 2015, 1507.05781.

[8]  Justin Solomon,et al.  Exponential Integration for Hamiltonian Monte Carlo , 2015, ICML.

[9]  Christian P. Robert,et al.  Bayesian computation: a summary of the current state, and samples backwards and forwards , 2015, Statistics and Computing.

[10]  Hongkai Zhao,et al.  Hamiltonian Monte Carlo Acceleration Using Neural Network Surrogate functions , 2015 .

[11]  Arthur Gretton,et al.  Gradient-free Hamiltonian Monte Carlo with Efficient Kernel Exponential Families , 2015, NIPS.

[12]  Robert Leenders,et al.  Hamiltonian ABC , 2015, UAI.

[13]  M. Betancourt,et al.  Optimizing The Integrator Step Size for Hamiltonian Monte Carlo , 2014, 1411.6669.

[14]  Ben Calderhead,et al.  A general construction for parallelizing Metropolis−Hastings algorithms , 2014, Proceedings of the National Academy of Sciences.

[15]  Tijana Radivojević,et al.  Constant pressure hybrid Monte Carlo simulations in GROMACS , 2014, Journal of Molecular Modeling.

[16]  M. Betancourt,et al.  The Geometric Foundations of Hamiltonian Monte Carlo , 2014, 1410.5110.

[17]  Jesús María Sanz-Serna,et al.  Extra Chance Generalized Hybrid Monte Carlo , 2014, J. Comput. Phys..

[18]  Jascha Sohl-Dickstein,et al.  Hamiltonian Monte Carlo Without Detailed Balance , 2014, ICML.

[19]  Yichuan Zhang,et al.  Semi-Separable Hamiltonian Monte Carlo for Inference in Bayesian Hierarchical Models , 2014, NIPS.

[20]  Jesús María Sanz-Serna,et al.  Numerical Integrators for the Hybrid Monte Carlo Method , 2014, SIAM J. Sci. Comput..

[21]  J. M. Sanz-Serna,et al.  Compressible generalized hybrid Monte Carlo. , 2014, The Journal of chemical physics.

[22]  Tianqi Chen,et al.  Stochastic Gradient Hamiltonian Monte Carlo , 2014, ICML.

[23]  Jan-Willem van de Meent,et al.  Tempering by Subsampling , 2014, 1401.7145.

[24]  M. Betancourt,et al.  Hamiltonian Monte Carlo for Hierarchical Models , 2013, 1312.0906.

[25]  Liam Paninski,et al.  Auxiliary-variable Exact Hamiltonian Monte Carlo Samplers for Binary Distributions , 2013, NIPS.

[26]  Babak Shahbaba,et al.  Spherical Hamiltonian Monte Carlo for Constrained Target Distributions , 2013, ICML.

[27]  T. Takaishi.  Empirical Analysis of Stochastic Volatility Model by Hybrid Monte Carlo Algorithm , 2013, 1305.3184.

[28]  Jitse Niesen,et al.  On an asymptotic method for computing the modified energy for symplectic methods , 2013, 1304.0673.

[29]  Nando de Freitas,et al.  Adaptive Hamiltonian and Riemann manifold Monte Carlo samplers , 2013, ICML 2013.

[30]  B. Escribano,et al.  Combining stochastic and deterministic approaches within high efficiency molecular simulations , 2013 .

[31]  Yichuan Zhang,et al.  Continuous Relaxations for Discrete Hamiltonian Monte Carlo , 2012, NIPS.

[32]  M. Girolami,et al.  Lagrangian Dynamical Monte Carlo , 2012, 1211.3759.

[33]  Jason A. Wagoner,et al.  Reducing the effect of Metropolization on mixing times in molecular dynamics simulations. , 2012, The Journal of chemical physics.

[34]  Jascha Sohl-Dickstein,et al.  Hamiltonian Monte Carlo with Reduced Momentum Flips , 2012, ArXiv.

[35]  Jascha Sohl-Dickstein,et al.  Hamiltonian Annealed Importance Sampling for partition function estimation , 2012, ArXiv.

[36]  Nando de Freitas,et al.  Adaptive MCMC with Bayesian Optimization , 2012, AISTATS.

[37]  Raquel Urtasun,et al.  A Family of MCMC Methods on Implicitly Defined Manifolds , 2012, AISTATS.

[38]  Babak Shahbaba,et al.  Split HMC for Gaussian Process Models , 2012, 1201.3973.

[39]  J. M. Sanz-Serna,et al.  Hybrid Monte Carlo on Hilbert spaces , 2011 .

[40]  Babak Shahbaba,et al.  Split Hamiltonian Monte Carlo , 2011, Stat. Comput..

[41]  M. Girolami,et al.  Riemann manifold Langevin and Hamiltonian Monte Carlo methods , 2011, Journal of the Royal Statistical Society: Series B (Statistical Methodology).

[42]  John K Kruschke,et al.  Bayesian data analysis. , 2010, Wiley interdisciplinary reviews. Cognitive science.

[43]  M. Betancourt.  Nested Sampling with Constrained Hamiltonian Monte Carlo , 2010, 1005.0157.

[44]  J. M. Sanz-Serna,et al.  Optimal tuning of the hybrid Monte Carlo algorithm , 2010, 1001.4460.

[45]  Andreas Krause,et al.  Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting , 2009, IEEE Transactions on Information Theory.

[46]  Scott S. Hampton,et al.  A separable shadow Hamiltonian hybrid Monte Carlo method. , 2009, The Journal of chemical physics.

[47]  Mark A. Girolami,et al.  Estimating Bayes factors via thermodynamic integration and population MCMC , 2009, Comput. Stat. Data Anal..

[48]  Nawaf Bou-Rabee,et al.  A comparison of generalized hybrid Monte Carlo methods with and without momentum flip , 2009, J. Comput. Phys..

[49]  A. Pettitt,et al.  Marginal likelihood estimation via power posteriors , 2008 .

[50]  Sebastian Reich,et al.  Improved sampling for simulations of interfacial membrane proteins: application of generalized shadow hybrid Monte Carlo to a peptide toxin/bilayer system. , 2008, The journal of physical chemistry. B.

[51]  Robert B. Gramacy,et al.  Importance tempering , 2007, Stat. Comput..

[52]  J. Rosenthal,et al.  Coupling and Ergodicity of Adaptive Markov Chain Monte Carlo Algorithms , 2007, Journal of Applied Probability.

[53]  David J. Earl,et al.  Parallel tempering: theory, applications, and new perspectives. , 2005, Physical chemistry chemical physics : PCCP.

[54]  Robert D. Skeel,et al.  Monitoring energy drift with shadow Hamiltonians , 2005 .

[55]  T. Takaishi,et al.  Testing and tuning symplectic integrators for the hybrid Monte Carlo algorithm in lattice QCD. , 2005, Physical review. E, Statistical, nonlinear, and soft matter physics.

[56]  Tim Hesterberg,et al.  Monte Carlo Strategies in Scientific Computing , 2002, Technometrics.

[57]  Robert D. Skeel,et al.  Practical Construction of Modified Hamiltonians , 2001, SIAM J. Sci. Comput..

[58]  Hoon Kim,et al.  Monte Carlo Statistical Methods , 2000, Technometrics.

[59]  A. Kennedy,et al.  Cost of the Generalised Hybrid Monte Carlo Algorithm for Free Field Theory , 2000, hep-lat/0008020.

[60]  Jun S. Liu,et al.  The Multiple-Try Method and Local Optimization in Metropolis Sampling , 2000 .

[61]  Y. Sugita,et al.  Replica-exchange molecular dynamics method for protein folding , 1999 .

[62]  Andrew Gelman,et al.  General methods for monitoring convergence of iterative simulations , 1998 .

[63]  Xiao-Li Meng,et al.  Simulating Normalizing Constants: From Importance Sampling to Bridge Sampling to Path Sampling , 1998 .

[64]  D. Rubin,et al.  Inference from Iterative Simulation Using Multiple Sequences , 1992 .

[65]  R. McLachlan,et al.  The accuracy of symplectic integrators , 1992 .

[66]  A. Horowitz.  A generalized guided Monte Carlo algorithm , 1991 .

[67]  T. Bayes.  An essay towards solving a problem in the doctrine of chances , 2003 .

[68]  Sourendu Gupta,et al.  The acceptance probability in the hybrid Monte Carlo method , 1990 .

[69]  Adrian F. M. Smith,et al.  Sampling-Based Approaches to Calculating Marginal Densities , 1990 .

[70]  Paul B. Mackenze.  An Improved Hybrid Monte Carlo Method , 1989 .

[71]  A. Kennedy,et al.  Hybrid Monte Carlo , 1988 .

[72]  Wang,et al.  Replica Monte Carlo simulation of spin glasses. , 1986, Physical review letters.

[73]  Donald Geman,et al.  Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images , 1984, IEEE Transactions on Pattern Analysis and Machine Intelligence.

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

[75]  Sebastian Reich,et al.  Multiple-time-stepping generalized hybrid Monte Carlo methods , 2015, J. Comput. Phys..

[76]  Ziyun Wang,et al.  Predictive Adaptation of Hybrid Monte Carlo with Bayesian Parametric Bandits , 2011 .

[77]  Nial Friel,et al.  Estimating the model evidence: a review , 2011 .

[78]  P. Atzberger.  The Monte-Carlo Method , 2006 .

[79]  Sebastian Reich,et al.  The Targeted Shadowing Hybrid Monte Carlo (TSHMC) Method , 2006 .

[80]  E. Hairer,et al.  Simulating Hamiltonian dynamics , 2006, Math. Comput..

[81]  Nando de Freitas,et al.  An Introduction to MCMC for Machine Learning , 2004, Machine Learning.

[82]  K. Hanson.  Use of probability gradients in hybrid MCMC and a new convergence test , 2002 .

[83]  A. Mira.  On Metropolis-Hastings algorithms with delayed rejection , 2001 .

[84]  Lingyu Chen,et al.  Exploring Hybrid Monte Carlo in Bayesian Computation , 2000 .

[85]  Jack J. Dongarra,et al.  Guest Editors Introduction to the top 10 algorithms , 2000, Comput. Sci. Eng..

[86]  P. Laplace.  Théorie analytique des probabilités , 1995 .

[87]  Robert I. McLachlan,et al.  On the Numerical Integration of Ordinary Differential Equations by Symmetric Composition Methods , 1995, SIAM J. Sci. Comput..

[88]  J. M. Sanz-Serna,et al.  Numerical Hamiltonian Problems , 1994 .

[89]  C. Geyer.  Markov Chain Monte Carlo Maximum Likelihood , 1991 .

[90]  John Geweke,et al.  Evaluating the accuracy of sampling-based approaches to the calculation of posterior moments , 1991 .

[91]  A. Kennedy.  The theory of hybrid stochastic algorithms , 1990 .

[92]  D. Frenkel.  Free-energy computation and first-order phase transitions , 1986 .

[93]  V. Arnold.  Mathematical Methods of Classical Mechanics , 1974 .

[94]  N. Metropolis.  THE BEGINNING of the MONTE CARLO METHOD , 2022 .