Randomized Halton sequences (Q1591883)

Monte Carlo and quasi-Monte Carlo methods are commonly used to evaluate multiple integrals. Quasi-Monte Carlo methods are based on careful choice of deterministic point sequences. These sequences called low-discrepancy sequences have nice distribution properties and provide faster convergence than classical Monte Carlo methods. On the other side low-discrepancy sequences have larger error bound. Examples of low discrepancy sequences are the Halton sequence, Sobol sequence, Faure sequence, Niederreiter sequence, and Niederreiter-Xing sequences. Some methods have been developed to generate low-discrepancy sequence with better error bound. They are based on ramdomization techniques of the sequences. This article describes a new method for randomizing the Halton sequence. This randomization makes use of the description of Halton sequence using von Neumann-Kakutani transformation. The starting point of the sequence is randomized. As mentioned before this method combines the potential accuracy advantage of Halton sequence in multi-dimension integration with the practical error estimation of Monte Carlo methods.