An extension of the digital method based on $b$-adic integers

Roswitha Hofer,Ísabel Pirsic
DOI: https://doi.org/10.48550/arXiv.1706.06323
2017-06-27
Abstract:We introduce a hybridization of digital sequences with uniformly distributed sequences in the domain of $b$-adic integers, $\mathbb Z_{b}, b\in\mathbb N\setminus\{1\}$, by using such sequences as input for generating matrices. The generating matrices are then naturally required to have finite row-lengths. We exhibit some relations of the `classical' digital method to our extended version, and also give several examples of new constructions with their respective quality assessments in terms of $t,\mathbf T$ and discrepancy.
Number Theory
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: in the digital sequence construction method, how to extend the classical digital method by using \(b\)-adic integers (i.e., \(Z_b\)) and ensure that the generated matrix has a finite row length. Specifically, the author introduced a hybrid method, combining uniformly distributed sequences with \(b\)-adic integers to generate new low - discrepancy sequences. ### Problem Background 1. **Importance of Low - Discrepancy Sequences** - Constructing sequences with good equidistribution properties is an important problem in number theory, especially in quasi - Monte Carlo methods, where star discrepancy is an important measure of uniform distribution. - Low - discrepancy sequences have a wide range of applications in numerical analysis, such as in integral calculation, optimization problems, etc. 2. **Limitations of Classical Digital Methods** - The classical digital method proposed by Niederreiter is usually used to generate low - discrepancy sequences, but its application scope and construction method have certain limitations. - The classical method mainly depends on non - negative integer sequences, and this paper aims to extend this method by introducing \(b\)-adic integers. ### Main Contributions of the Paper 1. **Extended Digital Method** - The author proposed a new algorithm (Algorithm 2) based on \(b\)-adic integers, which allows the use of any \(b\)-adic integer sequence as input, rather than being limited to non - negative integers. - In this way, more types of low - discrepancy sequences can be generated, thereby improving the efficiency and accuracy of the quasi - Monte Carlo method. 2. **Finite - Row Generation Matrix** - Ensure that each row of the generated matrix has only a finite number of non - zero elements, which guarantees the finiteness of the inner sum and makes the new algorithm more flexible and easier to implement. 3. **Theoretical Results and Examples** - The paper shows the relationship between the new method and the classical method and provides several new construction examples and their quality evaluations. - The effectiveness and superiority of the new method are verified through specific examples, especially in terms of star discrepancy and quality parameters \(t\), \(T\). ### Formula Summary - The star discrepancy is defined as: \[ D^*_N(P)=\sup_{J}\left|\frac{A(J;P)}{N}-\text{vol}(J)\right| \] where \(J\) is a sub - interval of \([0,1]^s\) and \(A(J;P)\) is the number of points falling in \(J\). - The \(i\)-th component in the new algorithm is defined as: \[ x^{(i)}_n=\sum_{j = 1}^{\infty}\lambda_{i,j}\left(\sum_{r = 0}^{\infty}c^{(i)}_{j,r}\psi_r(a_r)\right)/b^j \] where \(a_r\) is the coefficient in the \(b\)-adic representation of \(s_n\). Through these improvements, the paper provides a more general and effective low - discrepancy sequence generation method, which is suitable for a wider range of quasi - Monte Carlo application scenarios.