arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2102.02365math.NAcs.NAstat.APstat.ML
收藏

基于自适应随机傅立叶特征的风场重建

Wind Field Reconstruction with Adaptive Random Fourier Features

Jonas Kiessling, Emanuel Ström, Raúl Tempone

摘要:

我们调查使用空间插值方法重建水平近地面风场给定稀疏的测量。特别是,随机傅立叶特征相比,一组基准方法,包括克里格和反距离加权。随机傅立叶特征是近似速度场的线性模型$β(\pmb x) = \sum_{k=1}^K β_k e^{iω_k \pmb x}$,其中频率$ω_k$被随机采样并且幅度$β_k$被训练以使损失函数最小化。我们包括一个物理动机的发散惩罚项$|\nabla \cdot β(\pmb x)|^2$,以及Sobolev范数的惩罚。我们推导出的泛化误差的界限,并得出一个采样密度,最大限度地减少了边界。根据(arXiv:2007.10683 [math.NA]),我们设计了一种自适应Metropolis-Hastings算法,用于对最优分布的频率进行采样。在我们的实验中,我们的随机傅立叶特征模型优于基准模型。

英文摘要:

We investigate the use of spatial interpolation methods for reconstructing the horizontal near-surface wind field given a sparse set of measurements. In particular, random Fourier features is compared to a set of benchmark methods including Kriging and Inverse distance weighting. Random Fourier features is a linear model $β(\pmb x) = \sum_{k=1}^K β_k e^{iω_k \pmb x}$ approximating the velocity field, with frequencies $ω_k$ randomly sampled and amplitudes $β_k$ trained to minimize a loss function. We include a physically motivated divergence penalty term $|\nabla \cdot β(\pmb x)|^2$, as well as a penalty on the Sobolev norm. We derive a bound on the generalization error and derive a sampling density that minimizes the bound. Following (arXiv:2007.10683 [math.NA]), we devise an adaptive Metropolis-Hastings algorithm for sampling the frequencies of the optimal distribution. In our experiments, our random Fourier features model outperforms the benchmark models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑