The methods used for extracting an approximate eigenpair are crucial in sparse iterative eigensolvers. Using least squares and line search techniques this paper devises a method for an approximate eigenpair extraction. Numerical comparison of the Jacobi–Davidson method using the suggested method of eigenpair extraction, Rayleigh–Ritz, and refined Ritz projections shows that the suggested method is a viable alternative.

中文翻译：

---

最小二乘法和线搜索在 Jacobi-Davidson 方法中提取特征对

用于提取近似特征对的方法在稀疏迭代特征求解器中至关重要。本文使用最小二乘法和线搜索技术设计了一种近似特征对提取方法。使用建议的特征对提取方法、Rayleigh-Ritz 和精制 Ritz 投影对 Jacobi-Davidson 方法进行的数值比较表明，建议的方法是一种可行的替代方法。