Abstract:The main purpose of this work is the one of providing an efficient scheme for constructing reduced interpolation models for kernel bases. In literature such problem is mainly addressed via the well-established knot insertion or knot removal schemes. Such iterative strategies are usually quite demanding from a computational point of view and our goal is to study an efficient implementation for data removal approaches, namely Efficient Reduced Basis Algorithm (ERBA). Focusing on kernel-based interpolation, the algorithm makes use of two iterative rules for removing data. The former, called ERBA-r, is based on classical residual evaluations. The latter, namely ERBA-p, is independent of the function values and relies on error bounds defined by the power function. In both cases, inspired by the so-called extended Rippa's algorithm, our ERBA takes advantage of a fast implementation.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to provide an efficient method for constructing a kernel - function - based dimension - reduction interpolation model. Specifically, the author proposes a new method named "Efficient Reduced Basis Algorithm (ERBA)" to reduce the computational complexity and improve the efficiency.
### Main Problems
1. **Limitations of Existing Methods**
- Current reduced - basis methods mainly rely on traditional node insertion or deletion strategies, which are usually computationally expensive.
- Especially when dealing with large - scale data sets, these methods need to repeatedly construct multiple interpolation models, resulting in a huge consumption of computational resources.
2. **Requirement for Efficient Implementation**
- In order to overcome the problem of high computational complexity, researchers need a method that can significantly reduce the computational cost while ensuring the accuracy.
### Solutions
The ERBA proposed in the paper aims to solve the above problems in the following ways:
- **ERBA - r**: Based on the classical residual evaluation, by quickly implementing the Extended Rippa's Algorithm (ERA), it reduces the number of matrix inversions and linear system solutions, thereby reducing the computational complexity.
- **ERBA - p**: It does not depend on function values but is based on the error bound defined by the power function. By introducing an efficient power - function - vector - calculation method, it further simplifies the calculation process.
### Formula Summary
- The equation under the interpolation condition is:
\[
Ac = f
\]
where \( A=(A_{i,j}) = \kappa(x_i,x_j)\), \(i,j = 1,\ldots,n\), is the interpolation matrix.
- The power function is defined as:
\[
P_{\kappa,X}(x)=\sqrt{\kappa(x,x)-\kappa^{\top}(x)A^{-1}\kappa(x)}
\]
where \(\kappa(x)=(\kappa(x,x_1),\ldots,\kappa(x,x_n))^{\top}\).
- The residual calculation formula is:
\[
e_p=(A^{-1}_{p,p})^{-1}c_p
\]
- The power - function - vector calculation formula is:
\[
P_V=\sqrt{\text{sum}(((A^{-1}_{p,p})^{-1}A^{-1}_{p,:})\odot k(V)^{\top})}
\]
Through these improvements, ERBA not only improves the computational efficiency but also reduces the calculation time while maintaining the model accuracy, and is suitable for the efficient processing of large - scale data sets.