Abstract:This paper investigates the stability of the least squares approximation $P_m^n$ within the univariate polynomial space of degree $m$, denoted by ${\mathbb P}_m$. The approximation $P_m^n$ entails identifying a polynomial in ${\mathbb P}_m$ that approximates a function $f$ over a domain $X$ based on samples of $f$ taken at $n$ randomly selected points, according to a specified measure $\rho_X$. The primary goal is to determine the sampling rate necessary to ensure the stability of $P_m^n$. Assuming the sampling points are i.i.d. with respect to a Jacobi weight function, we present the sampling rates that guarantee the stability of $P_m^n$. Specifically, for uniform random sampling, we demonstrate that a sampling rate of $n \asymp m^2$ is required to maintain stability. By integrating these findings with those of Cohen-Davenport-Leviatan, we conclude that, for uniform random sampling, the optimal sampling rate for guaranteeing the stability of $P_m^n$ is $n \asymp m^2$, up to a $\log n$ factor. Motivated by this result, we extend the impossibility theorem, previously applicable to equally spaced samples, to the case of random samples, illustrating the balance between accuracy and stability in recovering analytic functions.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
The paper aims to study the stability of the least - squares approximation \(P_n^m\) under random sampling. Specifically, the paper focuses on how to determine the sampling rate that can ensure the stability of \(P_n^m\) in the polynomial space \(P_m\) when the sampling points are randomly selected from a given probability measure \(\rho_X\).
### Main problems and objectives
1. **Problem background**:
- Consider approximating an unknown function \(f\) by sampling the function \(f\) at \(n\) randomly selected points on an interval \(X = [- 1,1]\).
- The sampling points \(\{x_i\}_{i = 1}^n\) are independent and identically distributed (i.i.d.) and follow a certain probability measure \(\rho_X\).
- Use the discrete least - squares method \(P_n^m\) to find a polynomial \(p\in P_m\) such that \(p\) is as close as possible to \(f\) at these sampling points.
2. **Main objectives**:
- Determine the sampling rate required to ensure the stability of \(P_n^m\) under random sampling.
- In particular, when the sampling points follow the Jacobi weight function, give a specific sampling rate to ensure the stability of \(P_n^m\).
- For uniform random sampling, prove that the required sampling rate is \(n\asymp m^2\) and consider the influence of the logarithmic factor.
3. **Key problems**:
- **Problem I**: For uniform random sampling, can the random uniform sampling rate \(n\asymp m^2\) be reduced while maintaining high - probability stability?
- **Problem II**: In the case of random sampling, does the impossibility theorem still hold?
### Research contributions
1. **Answer to Problem I**:
- By analyzing the relationship between the stability of \(P_n^m\) and the sampling rate, it is proved that under the Jacobi weight function, the sampling rate \(n\asymp m^2(1 +\gamma)\) is necessary.
- For uniform random sampling (i.e., \(\gamma = 0\)), the sampling rate \(n\asymp m^2\) is a necessary condition to ensure the stability of \(P_n^m\).
2. **Answer to Problem II**:
- Extend the impossibility theorem to make it applicable to the case of random sampling.
- Prove the trade - off relationship between stability and accuracy under random sampling, similar to the results under deterministic sampling.
### Main conclusions
- **Theorem 1.4**: For random sampling under the Jacobi weight function, if the sampling rate is lower than \(n\asymp m^2(1 +\gamma)\), then exponential ill - conditioning will almost surely occur, so the sampling rate cannot be lower than \(n\asymp m^2(1 +\gamma)\).
- **Corollary 1.1**: Ignoring the logarithmic factor, the sampling rate \(n\asymp m^2(1 +\gamma)\) is a necessary and sufficient condition to ensure the high - probability stability of \(P_n^m\).
- **Theorem 1.5**: In the case of random sampling, the impossibility theorem still holds, indicating that with high probability, geometric convergence is necessarily accompanied by geometric instability.
Through these results, the paper provides a theoretical basis for understanding the stability of the least - squares approximation under random sampling and provides an important reference for research in related fields.