Independent Random Sampling Methods pp 115-157 | Cite as

# Adaptive Rejection Sampling Methods

## Abstract

This chapter is devoted to describing the class of the adaptive rejection sampling (ARS) schemes. These (theoretically) universal methods are very efficient samplers that update the proposal density whenever a generated sample is rejected in the RS test. In this way, they can produce i.i.d. samples from the target with an increasing acceptance rate that can converge to 1. As a by-product, these techniques also generate a sequence of proposal pdfs converging to the true shape of the target density. Another advantage of the ARS samplers is that, when they can be applied, the user only has to select a set of initial conditions. After the initialization, they are completely automatic, self-tuning algorithms (i.e., no parameters need to be adjusted by the user) regardless of the specific target density. However, the need to construct a suitable sequence of proposal densities restricts the practical applicability of this methodology. As a consequence, ARS schemes are often tailored to specific classes of target distributions. Indeed, the construction of the proposal is particularly hard in multidimensional spaces. Hence, ARS algorithms are usually designed only for drawing from univariate densities.

In this chapter we discuss the basic adaptive structure shared by all ARS algorithms. Then we look into the performance of the method, characterized by the acceptance probability (which increases as the proposal is adapted), and describe various extensions of the standard ARS approach which are aimed either at improving the efficiency of the method or at covering a broader class of target pdfs. Finally, we consider a hybrid method that combines the ARS and Metropolis-Hastings schemes.

## References

- 1.M.S. Arulumpalam, S. Maskell, N. Gordon, T. Klapp, A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking. IEEE Trans. Signal Process.
**50**(2), 174–188 (2002)Google Scholar - 2.J. Besag, P.J. Green, Spatial statistics and Bayesian computation. J. R. Stat. Soc. Ser. B
**55**(1), 25–37 (1993)Google Scholar - 3.C. Botts, A modified adaptive accept-reject algorithm for univariate densities with bounded support. Technical Report, http://williams.edu/Mathematics/cbotts/Research/paper3.pdf (2010)
- 4.G.E.P. Box, G.C. Tiao,
*Bayesian Inference in Statistical Analysis*(Wiley, New York, 1973)Google Scholar - 5.L. Devroye,
*Non-uniform Random Variate Generation*(Springer, New York, 1986)Google Scholar - 6.M. Evans, T. Swartz, Random variate generation using concavity properties of transformed densities. J. Comput. Graph. Stat.
**7**(4), 514–528 (1998)Google Scholar - 7.J. Geweke, Bayesian inference in econometric models using Monte Carlo integration. Econometrica
**24**, 1317–1399 (1989)Google Scholar - 8.W.R. Gilks, Derivative-free adaptive rejection sampling for Gibbs sampling. Bayesian Stat.
**4**, 641–649 (1992)Google Scholar - 9.W.R. Gilks, P. Wild, Adaptive rejection sampling for Gibbs sampling. Appl. Stat.
**41**(2), 337–348 (1992)Google Scholar - 10.W.R. Gilks, N.G. Best, K.K.C. Tan, Adaptive rejection Metropolis sampling within Gibbs sampling. Appl. Stat.
**44**(4), 455–472 (1995)Google Scholar - 11.D. Gorur, Y.W. Teh, Concave convex adaptive rejection sampling. J. Comput. Graph. Stat.
**20**(3), 670–691 (2011)Google Scholar - 12.T.L. Griffiths, Z. Ghahramani, The indian buffet process: an introduction and review. J. Mach. Learn. Res.
**12**, 1185–1224 (2011)Google Scholar - 13.H. Hirose, A. Todoroki, Random number generation for the generalized normal distribution using the modified adaptive rejection method. Int. Inf. Inst.
**8**(6), 829–836 (2005)Google Scholar - 14.L. Holden, Adaptive chains. Technical Report Norwegian Computing Center (1998)Google Scholar
- 15.L. Holden, R. Hauge, M. Holden, Adaptive independent Metropolis-Hastings. Ann. Appl. Probab.
**19**(1), 395–413 (2009)Google Scholar - 16.W. Hörmann, A rejection technique for sampling from T-concave distributions. ACM Trans. Math. Softw.
**21**(2), 182–193 (1995)Google Scholar - 17.W. Hörmann, J. Leydold, G. Derflinger,
*Automatic Nonuniform Random Variate Generation*(Springer, New York, 2003)Google Scholar - 18.W. Hörmann, J. Leydold, G. Derflinger, Inverse transformed density rejection for unbounded monotone densities. Research Report Series, Department of Statistics and Mathematics (Economy and Business), Vienna University (2007)Google Scholar
- 19.Y. Huang, J. Zhang, P.M. Djurić, Bayesian detection for BLAST. IEEE Trans. Signal Process.
**53**(3), 1086–1096 (2005)Google Scholar - 20.M.C. Jones, Distributions generated by transformation of scale using an extended cauchy-schlömilch transformation. Indian J. Stat.
**72-A**(2), 359–375 (2010)Google Scholar - 21.J. Leydold, J. Janka, W. Hörmann, Variants of transformed density rejection and correlation induction, in
*Monte Carlo and Quasi-Monte Carlo Methods 2000*(Springer, Heidelberg, 2002), pp. 345–356Google Scholar - 22.F. Liang, C. Liu, R. Caroll,
*Advanced Markov Chain Monte Carlo Methods: Learning from Past Samples*. Wiley Series in Computational Statistics (Wiley, Chichester, 2010)Google Scholar - 23.L. Martino,
*Novel Schemes for Adaptive Rejection Sampling*(Universidad Carlos III, Madrid, 2011)Google Scholar - 24.L. Martino, Parsimonious adaptive rejection sampling.
*IET Electron. Lett.***53**(16), 1115–1117 (2017)Google Scholar - 25.L. Martino, F. Louzada, Adaptive rejection sampling with fixed number of nodes. Commun. Stat. Simul. Comput., 1–11 (2017, to appear)Google Scholar
- 26.L. Martino, J. Míguez, A generalization of the adaptive rejection sampling algorithm. Stat. Comput.
**21**, 633–647 (2010). doi: https://doi.org/10.1007/s11222-010-9197-9 - 27.L. Martino, J. Míguez, Generalized rejection sampling schemes and applications in signal processing. Signal Process.
**90**(11), 2981–2995 (2010)Google Scholar - 28.L. Martino, J. Read, D. Luengo, Independent doubly adaptive rejection Metropolis sampling within Gibbs sampling,
*IEEE Trans. Signal Process.***63**(12), 3123–3138 (2015)Google Scholar - 29.L. Martino, R. Casarin, F. Leisen, D. Luengo, Adaptive independent sticky MCMC algorithms. EURASIP J. Adv. Signal Process. (2018, to appear)Google Scholar
- 30.R. Meyer, B. Cai, F. Perron, Adaptive rejection Metropolis sampling using Lagrange interpolation polynomials of degree 2. Comput. Stat. Data Anal.
**52**(7), 3408–3423 (2008)Google Scholar - 31.J. Michel, The use of free energy simulations as scoring functions. PhD Thesis, University of Southampton (2006)Google Scholar
- 32.R. Neal, MCMC using Hamiltonian dynamics, Chap. 5, in
*Handbook of Markov Chain Monte Carlo*ed. by S. Brooks, A. Gelman, G. Jones, X.-L. Meng (Chapman and Hall/CRC Press, Boca Raton, 2011)Google Scholar - 33.J.G. Proakis,
*Digital Communications*, 4th edn. (McGraw-Hill, Singapore, 2000)Google Scholar - 34.Y.W. Teh, D. Görür, Z. Ghahramani, Stick-breaking construction for the indian buffet process, in
*Proceedings of the International Conference on Artificial Intelligence and Statistics*(2007)Google Scholar - 35.L. Tierney, Exploring posterior distributions using Markov Chains, in
*Computer Science and Statistics: Proceedings of IEEE 23rd Symposium on the Interface*(1991), pp. 563–570Google Scholar