Abstract:Algebraic curve interpolation is described by specifying the location of N points in the plane and constructing an algebraic curve of a function f that should pass through them. In this paper, we propose a novel approach to construct the algebraic curve that interpolates a set of data (points or neighborhoods). This approach aims to search the polynomial with the smallest degree interpolating the given data. Moreover, the paper also presents an efficient method to reconstruct the algebraic curve of integer coefficients with the smallest degree and the least monomials that interpolates the provided data. The problems are converted into optimization problems and are solved via Lagrange multipliers methods and symbolic computation. Various examples are presented to illustrate the proposed approaches.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to construct an algebraic curve so that it can pass through a given set of data points or neighborhoods (such as square or spherical neighborhoods). Specifically, the author proposes a novel method to construct an interpolating polynomial with the lowest possible order and the simplest possible coefficients (i.e., integer coefficients with the smallest absolute value). In addition, this method also aims to solve the sparse interpolation problem, that is, to find the lowest - order polynomial curve with the fewest terms.
### Problem Description
The traditional polynomial interpolation problem usually seeks a function \(y = f(x_1, x_2,\dots, x_n)\) such that the deviation of this function at given points is minimized. However, the problem discussed in this paper is different: the given data contains small neighborhoods in space (such as square or spherical neighborhoods), and the goal is to construct an implicit polynomial function so that it passes through these small neighborhoods. The criterion for measuring the interpolation quality is not the deviation between the interpolation function and the known data, but the properties of the order and coefficients of the constructed polynomial. For example, the goal is to minimize the order of the polynomial and ensure that the coefficients are integers and their absolute values are as small as possible.
### Mathematical Representation
Given a set of \(N\) points \((x_i, y_i)\) and a positive integer \(d\), we define \(A[d]\) as a \(p\times p\) symmetric matrix, where \(p=\binom{d + 2}{2}\), and:
\[A[d]=\sum_{i = 1}^{N}A_i\]
where,
\[A_i=v_d(x_i, y_i)v_d(x_i, y_i)^T\]
Here, \(v_d(x_i, y_i)\) is a vector representing the values of the basis functions of the polynomial at the point \((x_i, y_i)\):
\[v_d(x_i, y_i)=[1, x_i, y_i, x_i^2, x_i y_i, y_i^2,\dots, x_i^{d - 1}y_i, y_i^d]^T\]
According to Theorem 2.2, if there is a curve \(f(x, y)=0\) passing through the given points, then:
\[\min_{f\in P}\text{degree}(f)\leq d\iff\det(A[d]) = 0\]
### Practical Applications
In practical applications, we often need to fit empirical formulas based on limited experimental or observational data. If the order of the interpolating polynomial is too high or the range of coefficients is too large, the predictive ability of the formula may be reduced. Therefore, it is hoped that the interpolation formula is as simple as possible (both the order and the coefficients are as small as possible) to avoid over - fitting and better capture the essential relationship between variables.
### Conclusion
This paper proposes a new method to solve the algebraic curve interpolation problem through optimization techniques (such as the Lagrange multiplier method and symbolic computation), especially when the given data contains small neighborhoods. This method can not only find the polynomial passing through these neighborhoods, but also ensure that the order and coefficients of the polynomial are as simple as possible, thereby improving the stability and practicality of the interpolation results.