STATISTICAL IMPROVEMENT SYSTEM FOR MONTE CARLO SIMULATION

摘要

PURPOSE: To reduce a statistical error by selecting Monte Carlo particles randomly as selected particles at a predetermined selection rate, removing unselected Monte Carlo particles as removed particles, multiplying the weight of the selected particles by a reciprocal number of the selection rate, and decreasing the number of Monte Carlo particles without varying the density distribution formed of the weight of the Monte Carlo particles. CONSTITUTION: When many Monte Carlo particles form some distribution, the Monte Carlo particles are selected randomly at the predetermined selection rate R and the weight (the number of carriers that one Monte Carlo particle handles) of the selected particles is multiplied by the reciprocal number of the selection rate R (1/R) to decrease the number of Monte Carlo particles without varying the density distribution formed of the weight of the Monte Carlo particles. Namely, (r) (0=r=1) is generated first with a uniform random number. When (r) is larger than the selection rate R, the particles are removed. When (r) is less than R, the particles are selected as selected particles and the weight of the selected particles is multiplied by 1/R.