Dionisio F. Yáñez
Abstract:In this work, we present some new integration formulas for any order of accuracy as an application of the B-spline relations obtained in [1]. The resulting rules are defined as a perturbation of the trapezoidal integration method. We prove the order of approximation and extend the results to several dimensions. Finally, some numerical experiments are performed in order to check the theoretical results.
What problem does this paper attempt to address?
This paper aims to solve the problem of improving the integration accuracy in numerical integration. Specifically, the author proposes a new integration formula based on B - spline basis, and these formulas can achieve the accuracy of any order. The new method is an improvement or modification of the traditional trapezoidal integration rule, achieving higher accuracy by introducing B - spline relationships.
### Main problems solved in the paper:
1. **Improving the accuracy of numerical integration**: The paper proposes new integration formulas that can achieve the accuracy of any order, which is achieved by modifying the traditional trapezoidal integration rule.
2. **Extension to the multi - dimensional case**: The author not only verifies the effectiveness of the new formula in the single - variable case, but also extends it to the multi - dimensional case, using the tensor product strategy to deal with multi - dimensional integration problems.
3. **Theoretical and experimental verification**: The paper not only theoretically proves the order and accuracy of the new formula, but also verifies these theoretical results through numerical experiments.
### Specific content:
- **Background introduction**: The paper first reviews the basic concepts of numerical integration, especially the trapezoidal integration rule, and introduces B - spline functions and their applications in numerical integration.
- **Derivation of new formulas**: The author uses B - spline basis functions and quasi - interpolation operators to derive a series of new integration formulas. These formulas can be regarded as modified versions of the trapezoidal integration rule.
- **Theoretical analysis**: The paper proves in detail the order and accuracy of the new formulas and shows the connection between these formulas and the Euler - McClaurin formula.
- **Numerical experiments**: Through specific numerical experiments, the effectiveness and accuracy of the new formulas are verified, especially in some cases, the new formulas can achieve higher accuracy with fewer function evaluations.
### Conclusion:
- The newly proposed integration formulas can achieve higher accuracy at a lower computational cost in some cases.
- These formulas are flexible and simple and are suitable for a variety of numerical integration problems.
- Future work is recommended to further optimize these formulas, for example, using the hierarchical basis strategy to reduce the number of function evaluations.
### Key formulas:
- **Trapezoidal integration rule**:
\[
T_{[a,b],N}(f)=\frac{h_N}{2}(f(x_0)+f(x_N))+h_N\sum_{i = 1}^{N - 1}f(x_i)
\]
where \(h_N=\frac{b - a}{N}\).
- **New integration formula**:
\[
T_p^{[a,b],N}(f)=T_{[a,b],N}(f)+h\sum_{i = 1}^{2\lfloor p/2\rfloor}\xi_{p,-i}(f(x_{-i})-f(x_i)+f(x_{N + i})-f(x_{N - i}))
\]
- **B - integration term**:
\[
\tau_{p,j}=\sum_{r = j - 1-\lfloor p/2\rfloor}^{j+\lfloor p/2\rfloor}c_{p,r}B_{p + 1}\left(\frac{r - j + 1}{2}\right)
\]
These formulas and methods provide a new, flexible and efficient tool for numerical integration.