Performance analysis of random sampling algorithms on GNU Octave

International Nuclear Science, Technology and Engineering Conference 2018

Bin Yazid, Mohd Syahmi^1 ; Lemaire, Matthieu^2 ; Mohd Sies, Mohsin Bin^1 ; Lee, Deokjung^2

School of Chemical and Energy Engineering, Faculty of Engineering, Universiti Teknologi Malaysia, Johor Bahru

81310, Malaysia^1

Department of Nuclear Engineering, Ulsan National Institute of Science and Technology, 50 UNIST-gil, Ulsan

44919, Korea, Republic of^2

关键词: Constant functions;    Generation time;    Inverse transform method;    Monte Carlo codes;    Performance analysis;    Photon transport;    Random sampling;    Random sampling method;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/555/1/012003/pdf
DOI  :  10.1088/1757-899X/555/1/012003

来源: IOP

【 摘 要 】

This paper presents the performance analysis of two random sampling algorithms, the inverse-transform method and the Vose aliasing method, on GNU Octave. The Monte Carlo code MCS developed by UNIST uses random sampling methods to simulate the physics of neutron and photon transport [1]. The goal is to optimize the sampling time of MCS for cases when the probability density function is a constant function throughout the simulation. For this purpose, the runtime of the inverse-transform method and Vose aliasing method are compared for increasing input size with scripts developed on GNU Octave. To compare the execution time, the initialization and generation time of both methods are determined and discussed.