Self-Test Monte Carlo Method
スポンサーリンク
概要
- 論文の詳細を見る
The Self-Test Monte Carlo (STMC) method resolves the main problems in using algebraic pseudo-random numbers for Monte Carlo (MC) calculations: that they can interfere with MC algorithms and lead to erroneous results, and that such an error often cannot be detected without known exact solution. STMC is based on good randomness of about 10^<10> bits available from physical noise or transcendental numbers like π=3.14・・・. Various bit modifiers are available to get more bits for applications that demands more than 10^<10> random bits such as lattice quantum chromodynamics (QCD). These modifiers are designed so that a) each of them gives a bit sequence comparable in randomness as the original if used separately from each other, and b) their mutual interference when used jointly in a single MC calculation is adjustable. Intermediate data of the MC calculation itself are used to quantitatively test and adjust the mutual interference of the modifiers in respect of the MC algorithm. STMC is free of systematic error and gives reliable statistical error. Also it can be easily implemented on vector and parallel supercomputers.
- 理論物理学刊行会の論文
- 1996-07-25