Adaptive Monte Carlo algorithm for Wigner kernel evaluation

Abstract

In this paper, we study numerically various approaches, namely an adaptive Monte Carlo algorithm, a particular rank-1 lattice algorithm based on generalized Fibonacci numbers and a Monte Carlo algorithm based on Latin hypercube sampling for computing multidimensional integrals. We compare the performance of the algorithms over three case studies—multidimensional integrals from Bayesian statistics, the so-called Genz test functions and the Wigner kernel—an important issue in quantum mechanics represented by multidimensional integrals. A comprehensive study and an analysis of the computational complexity of the algorithms under consideration has been presented. Adaptive strategy is well-established as an efficient and reliable tool for multidimensional integration of integrands functions with computational peculiarities like peaks. The presented adaptive Monte Carlo algorithm gives reliable results in computing the Wigner kernel by a stochastic approach that has significantly lower computational complexity than the existing deterministic approaches.