Abstract:We study a random sampling technique to approximate integrals $\int_{[0,1]^s}f(\mathbf{x})\,\mathrm{d}\mathbf{x}$ by averaging the function at some sampling points. We focus on cases where the integrand is smooth, which is a problem which occurs in statistics. The convergence rate of the approximation error depends on the smoothness of the function $f$ and the sampling technique. For instance, Monte Carlo (MC) sampling yields a convergence of the root mean square error (RMSE) of order $N^{-1/2}$ (where $N$ is the number of samples) for functions $f$ with finite variance. Randomized QMC (RQMC), a combination of MC and quasi-Monte Carlo (QMC), achieves a RMSE of order $N^{-3/2+\varepsilon}$ under the stronger assumption that the integrand has bounded variation. A combination of RQMC with local antithetic sampling achieves a convergence of the RMSE of order $N^{-3/2-1/s+\varepsilon}$ (where $s\ge1$ is the dimension) for functions with mixed partial derivatives up to order two. Additional smoothness of the integrand does not improve the rate of convergence of these algorithms in general. On the other hand, it is known that without additional smoothness of the integrand it is not possible to improve the convergence rate. This paper introduces a new RQMC algorithm, for which we prove that it achieves a convergence of the root mean square error (RMSE) of order $N^{-\alpha-1/2+\varepsilon}$ provided the integrand satisfies the strong assumption that it has square integrable partial mixed derivatives up to order $\alpha>1$ in each variable. Known lower bounds on the RMSE show that this rate of convergence cannot be improved in general for integrands with this smoothness. We provide numerical examples for which the RMSE converges approximately with order $N^{-5/2}$ and $N^{-7/2}$, in accordance with the theoretical upper bound.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to increase the convergence rate of numerical integration, especially for the case of smooth integrands. Specifically:
- **Background problem**: In fields such as statistics, it is often necessary to calculate multi - dimensional integrals \(\int_{[0,1]^s} f(x) \, dx\). When dealing with such integrals, the convergence rate of the root mean square error (RMSE) of traditional Monte Carlo (MC) methods and quasi - Monte Carlo (QMC) methods depends on the smoothness of the integrand \(f\) and the sampling technique.
- For a function \(f\) with finite variance, the RMSE convergence rate of the Monte Carlo method is \(O(N^{-1/2})\).
- For a function \(f\) with bounded variation, the RMSE convergence rate of the randomized quasi - Monte Carlo (RQMC) method is \(O(N^{-3/2+\epsilon})\).
- For the RQMC method combined with local antithetic sampling, for a function \(f\) with second - order mixed partial derivatives, the RMSE convergence rate is \(O(N^{-3/2 - 1/s+\epsilon})\).
- **Limitations of existing methods**: Although the above methods perform well in some cases, for functions with higher - order smoothness, the convergence rates of these methods do not improve significantly. That is, when the integrand \(f\) has higher - order square - integrable mixed partial derivatives, the existing RQMC methods cannot fully utilize this additional smoothness to further improve the convergence rate.
- **Core problem of the paper**: This paper aims to introduce a new randomized quasi - Monte Carlo algorithm so that when the integrand \(f\) has square - integrable \(\alpha\)-order mixed partial derivatives in each variable, the convergence rate of RMSE can reach \(O(N^{-\alpha - 1/2+\epsilon})\). This convergence rate is the best possible result known so far and is applicable for any \(\alpha>0\).
By introducing this new algorithm, the author hopes to fully utilize the high - order smoothness of the integrand, thereby achieving a faster convergence rate, which is especially important when dealing with high - dimensional integral problems.