Abstract:This work considers parameter estimation for Gaussian process interpolation with a periodized version of the Mat{é}rn covariance function introduced by Stein. Convergence rates are studied for the joint maximum likelihood estimation of the regularity and the amplitude parameters when the data are sampled according to the model. The mean integrated squared error is also analyzed with fixed and estimated parameters, showing that maximum likelihood estimation yields asymptotically the same error as if the ground truth was known. Finally, the case where the observed function is a fixed deterministic element of a Sobolev space of continuous functions is also considered, suggesting that a joint estimation does not select the regularity parameter as if the amplitude were fixed.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is about the asymptotic properties of Matérn model parameter estimation in Gaussian process interpolation. Specifically, the author studies the convergence rate of the joint maximum - likelihood estimation of the regularity and amplitude parameters when data are sampled according to the model. In addition, the mean - squared error under fixed and estimated parameters is also analyzed, and it is explored whether the joint estimation will select the regularity parameter when the observed function is a fixed deterministic element in the continuous function space, as in the case where the amplitude parameter is fixed.
### Main Problem Analysis
1. **Asymptotic Properties of Parameter Estimation**:
- The paper focuses on how to estimate the asymptotic properties of the regularity parameter \(\nu\) and the amplitude parameter \(\phi\) in the Matérn model by the maximum - likelihood estimation method.
- The author studies the convergence rates of these parameter estimations under different conditions, especially the asymptotic properties of the maximum - likelihood estimation in the fixed - domain framework (i.e., within a bounded domain).
2. **Analysis of Mean - Squared Error**:
- The author analyzes the Mean Integrated Squared Error (MISE) under fixed and estimated parameters and shows that the maximum - likelihood estimation has the same error in the asymptotic case as in the case with known true parameters.
3. **Case of Fixed Deterministic Function**:
- The paper also considers whether the joint estimation will select the regularity parameter when the observed function is a fixed deterministic element in the Sobolev space, as in the case where the amplitude parameter is fixed.
- The author finds that in this case, the joint estimation will not select the regularity parameter as in the case where the amplitude parameter is fixed, but has a smaller asymptotic lower bound \(\liminf \hat{\nu}_n\geq \nu_0(f)-\frac{1}{2}\).
### Formula Summary
- **Spectral Representation of Matérn Covariance Function**:
\[
\hat{k}(\omega)=\frac{\phi}{(\alpha^2 + \|\omega\|^2)^{\nu + d/2}}
\]
where \(\nu\) is the regularity parameter, \(\phi\) is the amplitude parameter, and \(\alpha\) is the scale parameter.
- **Fourier Series Representation of Periodized Matérn Covariance Function**:
\[
k_\theta(x)=\sum_{j\in\mathbb{Z}} c_j(\theta) e^{2\pi i x j}
\]
where
\[
c_j(\theta)=\frac{\phi}{(\alpha^2 + j^2)^{\nu + 1/2}}
\]
- **Log - Likelihood Function**:
\[
L_n(\theta)=\frac{1}{n}(\ln(\det(K_\theta))+Z^T K_\theta^{-1} Z)
\]
- **Profile Likelihood Function**:
\[
\inf_{\phi > 0} L_n(\nu,\phi,\alpha)=1+\frac{1}{n}\ln(\det(R_{\nu,\alpha}))+\ln\left(\frac{Z^T R_{\nu,\alpha}^{-1} Z}{n}\right)
\]
- **Asymptotic Lower Bound**:
\[
\liminf \hat{\nu}_n\geq \nu_0(f)-\frac{1}{2}
\]
### Conclusion
The main contribution of the paper is to prove the strong consistency and asymptotic normality of the maximum - likelihood estimations of the regularity parameter and the amplitude parameter, and to analyze the performance of these estimations on prediction errors. In addition, the paper also explores the selection behavior of joint estimations in the case of fixed deterministic functions, providing some new insights.