Subspace iterations are used to minimise a generalised Ritz functional of alarge, sparse Hermitean matrix. In this way, the lowest $m$ eigenvalues aredetermined. Tests with $1 \leq m \leq 32$ demonstrate that the computationalcost (no. of matrix multiplies) does not increase substantially with $m$. Thisimplies that, as compared to the case of a $m=1$, the additional eigenvaluesare obtained for free.
展开▼
机译：子空间迭代用于最大程度地减少大型稀疏埃尔米特矩阵的广义Ritz函数。这样，确定了最低的$ m $特征值。用$ 1 \ leq m \ leq 32 $进行的测试表明，$ m $不会大大增加计算成本（矩阵乘法的数量）。这意味着，与$ m = 1 $的情况相比，可以免费获得其他特征值。
展开▼
