Abstract:Higher order scrambled digital nets are randomized quasi-Monte Carlo rules which have recently been introduced in [J. Dick, Ann. Statist., 39 (2011), 1372--1398] and shown to achieve the optimal rate of convergence of the root mean square error for numerical integration of smooth functions defined on the $s$-dimensional unit cube. The key ingredient there is a digit interlacing function applied to the components of a randomly scrambled digital net whose number of components is $ds$, where the integer $d$ is the so-called interlacing factor. In this paper, we replace the randomly scrambled digital nets by randomly scrambled polynomial lattice point sets, which allows us to obtain a better dependence on the dimension while still achieving the optimal rate of convergence. Our results apply to Owen's full scrambling scheme as well as the simplifications studied by Hickernell, Matoušek and Owen. We consider weighted function spaces with general weights, whose elements have square integrable partial mixed derivatives of order up to $\alpha\ge 1$, and derive an upper bound on the variance of the estimator for higher order scrambled polynomial lattice rules. Employing our obtained bound as a quality criterion, we prove that the component-by-component construction can be used to obtain explicit constructions of good polynomial lattice point sets. By first constructing classical polynomial lattice point sets in base $b$ and dimension $ds$, to which we then apply the interlacing scheme of order $d$, we obtain a construction cost of the algorithm of order $\mathcal{O}(dsmb^m)$ operations using $\mathcal{O}(b^m)$ memory in case of product weights, where $b^m$ is the number of points in the polynomial lattice point set.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to improve the numerical calculation accuracy and efficiency of high - dimensional integrals, especially for smooth functions defined on the s - dimensional unit cube. Specifically, the authors aim to achieve this goal by constructing interlaced scrambled polynomial lattice rules. This method can maintain the advantages of the quasi - Monte Carlo (QMC) method, that is, the selected quadrature points are distributed as evenly as possible, and at the same time, randomize these points to obtain better convergence and error estimates.
### Core of the problem
1. **Improve the convergence rate**: Traditional digital nets and sequences have been proven to be useful for efficiently performing high - dimensional integrals, but for higher - order smooth functions, the convergence rate needs to be further optimized. The method proposed in this paper can achieve the optimal root - mean - square error (RMSE) convergence rate \(N^{-(2\min(\alpha,d)+1)+\delta}\), where \(N\) is the number of sampling points, \(\alpha\) is the smoothness of the function, and \(d\) is the interlacing factor.
2. **Reduce dimension dependence**: When dealing with high - dimensional problems, many existing QMC methods have a strong dependence on the dimension, resulting in a significant performance degradation as the dimension increases. By using interlaced scrambled polynomial lattice rules, this paper can reduce the dependence on the dimension while maintaining the optimal convergence rate.
3. **Explicit construction of high - quality point sets**: To ensure that the constructed point sets have good properties, the author introduced the component - by - component (CBC) construction algorithm and proved that this algorithm can find polynomial lattice rules that meet specific quality standards.
### Specific problem description
- **Multivariable integration problem**: Study how to approximate multivariable integrals by averaging function values
\[
I(f)=\int_{[0,1]^s} f(x) \, dx,
\]
using \(N\) points \(x_0, x_1,\ldots, x_{N - 1}\) for numerical integration:
\[
\hat{I}(f)=\frac{1}{N}\sum_{n = 0}^{N - 1} f(x_n).
\]
- **Randomized QMC rules**: By randomizing the deterministic quadrature points (such as Owen's scrambling algorithm), the basic structure is preserved, which not only simplifies the error estimate but also improves the convergence rate.
- **Higher - order smooth functions**: For functions with square - integrable mixed partial derivatives (smoothness \(\alpha\geq1\)), construct higher - order digital nets or sequences through the interlacing algorithm to further improve the convergence rate of the integral error variance.
### Solution
The authors proposed the interlaced scrambled polynomial lattice rules and achieved the above goals through the following steps:
1. **Construct the classical polynomial lattice rules**: First, construct the classical polynomial lattice rules in dimension \(ds\).
2. **Apply Owen's scrambling algorithm**: Randomize the quadrature points of the polynomial lattice rules.
3. **Apply the interlacing algorithm**: Convert the scrambled points into a randomized quadrature rule applicable to the \(s\)-dimensional space through the interlacing algorithm.
In this way, the authors not only obtained a higher convergence rate, but also reduced the dependence on the dimension, and provided an explicit construction method.