On zero variance Monte Carlo path-stretching schemes

Lux, I

Title: On zero variance Monte Carlo path-stretching schemes

Full Record
Other Related Research

Abstract

A zero variance path-stretching biasing scheme proposed for a special case by Dwivedi is derived in full generality. The procedure turns out to be the generalization of the exponential transform. It is shown that the biased game can be interpreted as an analog simulation procedure, thus saving some computational effort in comparison with the corresponding nonanalog game.

Authors:: Lux, I

Publication Date:: Mon Aug 01 00:00:00 EDT 1983

Research Org.:: Central Research Institute for Physics, P.O. Box 49, 1525 Budapest

OSTI Identifier:: 5313847

Resource Type:: Journal Article

Journal Name:: Nucl. Sci. Eng.; (United States)

Additional Journal Information:: Journal Volume: 84:4

Country of Publication:: United States

Language:: English

Subject:: 22 GENERAL STUDIES OF NUCLEAR REACTORS; NEUTRON TRANSPORT THEORY; MONTE CARLO METHOD; REACTOR KINETICS; MIGRATION LENGTH; REACTOR SIMULATORS; SIMULATION; ANALOG SYSTEMS; DIMENSIONS; FUNCTIONAL MODELS; KINETICS; LENGTH; SIMULATORS; TRANSPORT THEORY; 220100* - Nuclear Reactor Technology- Theory & Calculation

Citation Formats


                    Lux, I. On zero variance Monte Carlo path-stretching schemes.  United States: N. p., 1983. 
        Web.

Copy to clipboard


                    Lux, I. On zero variance Monte Carlo path-stretching schemes.  United States.

Copy to clipboard


                    Lux, I. 1983.  
        "On zero variance Monte Carlo path-stretching schemes".  United States.

Copy to clipboard


                    
@article{osti_5313847,

  title        = {On zero variance Monte Carlo path-stretching schemes},

  author       = {Lux, I},

  abstractNote = {A zero variance path-stretching biasing scheme proposed for a special case by Dwivedi is derived in full generality. The procedure turns out to be the generalization of the exponential transform. It is shown that the biased game can be interpreted as an analog simulation procedure, thus saving some computational effort in comparison with the corresponding nonanalog game.},

  doi          = {},

  url          = {https://www.osti.gov/biblio/5313847},
  journal      = {Nucl. Sci. Eng.; (United States)},
number       = ,

  volume       = 84:4,

  place        = {United States},

  year         = {Mon Aug 01 00:00:00 EDT 1983},

  month        = {Mon Aug 01 00:00:00 EDT 1983}

}

Copy to clipboard

Journal Article:

Other availability

Find in Google Scholar

Search WorldCat to find libraries that may hold this journal

Save / Share:

Export Metadata

Save to My Library

Similar records in OSTI.GOV collections:

A class of zero-variance biasing schemes for Monte Carlo reaction rate estimators

Journal Article Gupta, M - Nucl. Sci. Eng.; (United States)

The radiation transport kernel has two factors, namely, the transition kernel and the collision kernel. Using the modified moments equations for analog and nonanalog games, families of zero-variance biasing schemes have been obtained for the two factors for evaluating reaction rate integrals using last-event, collision, and partial score estimators. It is shown that all zero-variance biasing schemes available in the literature belong to the respective family for each estimator.
Implementation of an approximate zero-variance scheme in the TRIPOLI Monte Carlo code

Conference Christoforou, S; Hoogenboom, J; Dumonteil, E; ...

In an accompanying paper it is shown that theoretically a zero-variance Monte Carlo scheme can be devised for criticality calculations if the space, energy and direction dependent adjoint function is exactly known. This requires biasing of the transition and collision kernels with the appropriate adjoint function. In this paper it is discussed how an existing general purpose Monte Carlo code like TRIPOLI can be modified to approach the zero-variance scheme. This requires modifications for reading in the adjoint function obtained from a separate deterministic calculation for a number of space intervals, energy groups and discrete directions. Furthermore, a function hasmore »« less
Deterministic theory of Monte Carlo variance

Journal Article Ueki, T; Larsen, E - Transactions of the American Nuclear Society

The theoretical estimation of variance in Monte Carlo transport simulations, particularly those using variance reduction techniques, is a substantially unsolved problem. In this paper, the authors describe a theory that predicts the variance in a variance reduction method proposed by Dwivedi. Dwivedi`s method combines the exponential transform with angular biasing. The key element of this theory is a new modified transport problem, containing the Monte Carlo weight w as an extra independent variable, which simulates Dwivedi`s Monte Carlo scheme. The (deterministic) solution of this modified transport problem yields an expression for the variance. The authors give computational results that validatemore » this theory.« less
A zero-variance-based scheme for variance reduction in Monte Carlo criticality

Conference Christoforou, S; Hoogenboom, J

A zero-variance scheme is derived and proven theoretically for criticality cases, and a simplified transport model is used for numerical demonstration. It is shown in practice that by appropriate biasing of the transition and collision kernels, a significant reduction in variance can be achieved. This is done using the adjoint forms of the emission and collision densities, obtained from a deterministic calculation, according to the zero-variance scheme. By using an appropriate algorithm, the figure of merit of the simulation increases by up to a factor of 50, with the possibility of an even larger improvement. In addition, it is shownmore »« less
Simultaneous Monte Carlo zero-variance estimates of several correlated means

Journal Article Booth, T - Transactions of the American Nuclear Society

Zero-variance procedures have been in existence since the dawn of Monte Carlo. Previous works all treat the problem of zero-variance solutions for a single tally. However, suppose one is interested in a whole series of related tallies? For example, suppose one desires the current exiting a shield in each of 1000 angular bins? A zero-variance solution for any angle bin i is an infinite variance solution in every angle bin j {ne} i, yet the zero-variance biasing required for all angle bins is very similar until within about a free path of the tally surface. This paper exploits this similaritymore »« less

Similar Records