Massively Parallel Computation Using Graphics Processors with Application to Optimal Experimentation in Dynamic Control
The rapid increase in the performance of graphics hardware, coupled with recent improvements in its programmability has lead to its adoption in many non-graphics applications, including wide variety of scientific computing fields. At the same time, a number of important dynamic optimal policy problems in economics are athirst of computing power to help overcome dual curses of complexity and dimensionality. We investigate if computational economics may benefit from new tools on a case study of imperfect information dynamic programming problem with learning and experimentation trade-off that is, a choice between controlling the policy target and learning system parameters. Specifically, we use a model of active learning and control of linear autoregression with unknown slope that appeared in a variety of macroeconomic policy and other contexts. The endogeneity of posterior beliefs makes the problem difficult in that the value function need not be convex and policy function need not be continuous. This complication makes the problem a suitable target for massively-parallel computation using graphics processors. Our findings are cautiously optimistic in that new tools let us easily achieve a factor of 15 performance gain relative to an implementation targeting single-core processors and thus establish a better reference point on the computational speed vs. coding complexity trade-off frontier. While further gains and wider applicability may lie behind steep learning barrier, we argue that the future of many computations belong to parallel algorithms anyway.
|Date of creation:||09 Jul 2009|
|Date of revision:|
|Contact details of provider:|| Postal: Ludwigstraße 33, D-80539 Munich, Germany|
Web page: https://mpra.ub.uni-muenchen.de
More information through EDIRC
Please report citation or reference errors to , or , if you are the registered author of the cited work, log in to your RePEc Author Service profile, click on "citations" and make appropriate adjustments.:
- Wilbur John Coleman II, 1992.
"Solving nonlinear dynamic models on parallel computers,"
Discussion Paper / Institute for Empirical Macroeconomics
66, Federal Reserve Bank of Minneapolis.
- Coleman, Wilbur John, II, 1993. "Solving Nonlinear Dynamic Models on Parallel Computers," Journal of Business & Economic Statistics, American Statistical Association, vol. 11(3), pages 325-30, July.
- Yock Y. Chong & David F. Hendry, 1986. "Econometric Evaluation of Linear Macro-Economic Models," Review of Economic Studies, Oxford University Press, vol. 53(4), pages 671-690.
- Goffe, William L. & Ferrier, Gary D. & Rogers, John, 1994. "Global optimization of statistical functions with simulated annealing," Journal of Econometrics, Elsevier, vol. 60(1-2), pages 65-99.
- Prescott, Edward C, 1972. "The Multi-Period Control Problem Under Uncertainty," Econometrica, Econometric Society, vol. 40(6), pages 1043-58, November.
- Beck, Gunter W. & Wieland, Volker, 2002. "Learning and control in a changing economic environment," Journal of Economic Dynamics and Control, Elsevier, vol. 26(9-10), pages 1359-1377, August.
- Sims, Christopher A. & Waggoner, Daniel F. & Zha, Tao, 2008.
"Methods for inference in large multiple-equation Markov-switching models,"
Journal of Econometrics,
Elsevier, vol. 146(2), pages 255-274, October.
- Christopher A. Sims & Daniel F. Waggoner & Tao Zha, 2006. "Methods for inference in large multiple-equation Markov-switching models," FRB Atlanta Working Paper 2006-22, Federal Reserve Bank of Atlanta.
- Nagurney, Anna & Takayama, Takashi & Zhang, Ding, 1995. "Massively parallel computation of spatial price equilibrium problems as dynamical systems," Journal of Economic Dynamics and Control, Elsevier, vol. 19(1-2), pages 3-37.
- David Hendry & Neil Shephard & Jurgen Doornik, 2001.
"Computationally-intensive Econometrics using a Distributed Matrix-programming Language,"
Economics Series Working Papers
2001-W22, University of Oxford, Department of Economics.
- Jurgen A. Doornik & David F. Hendry & Neil Shephard, . "Computationally-intensive Econometrics using a Distributed Matrix-programming Language," Economics Papers 2001-W22, Economics Group, Nuffield College, University of Oxford.
- Christopher Ferrall, 2003. "Solving Finite Mixture Models in Parallel," Computational Economics 0303003, EconWPA.
- Michael Creel, 2005. "User-Friendly Parallel Computations with Econometric Examples," Computational Economics, Society for Computational Economics, vol. 26(2), pages 107-128, October.
- Brezzi, Monica & Lai, Tze Leung, 2002. "Optimal learning and experimentation in bandit problems," Journal of Economic Dynamics and Control, Elsevier, vol. 27(1), pages 87-108, November.
- Wieland, Volker, 2000.
"Learning by doing and the value of optimal experimentation,"
Journal of Economic Dynamics and Control,
Elsevier, vol. 24(4), pages 501-534, April.
- Volker W. Wieland, 1996. "Learning by doing and the value of optimal experimentation," Finance and Economics Discussion Series 96-5, Board of Governors of the Federal Reserve System (U.S.).
- Swann, Christopher A, 2002. "Maximum Likelihood Estimation Using Parallel Computing: An Introduction to MPI," Computational Economics, Society for Computational Economics, vol. 19(2), pages 145-78, April.
- Kendrick, David, 1978. "Non-convexities from probing in adaptive control problems," Economics Letters, Elsevier, vol. 1(4), pages 347-351.
- Michael Creel, 2005. "User-Friendly Parallel Computations with Econometric Examples," UFAE and IAE Working Papers 637.05, Unitat de Fonaments de l'Anàlisi Econòmica (UAB) and Institut d'Anàlisi Econòmica (CSIC).
When requesting a correction, please mention this item's handle: RePEc:pra:mprapa:16721. See general information about how to correct material in RePEc.
For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: (Joachim Winter)
If references are entirely missing, you can add them using this form.