LanczosEigenSolver() returns an instance of this class which gives the information of the Lanczos procedure. More...
#include <laneig.h>
Public Attributes | |
| mkIndex | numIter |
| Number of iterations performed in the Lanczos procedure. | |
| Real | forNextKrylovVectorCpuTime |
| CPU time (in total) for obtaining the next Krylov vector in each iteration, e.g. matrix-vector products. | |
| Real | reorthogonalizationCpuTime |
| CPU time (in total) for partial or full reorthogonalization. | |
| Real | convergenceCheckCpuTime |
| CPU time (in total) for convergence check. | |
| Real | forEigenVectorsCpuTime |
| CPU time for obtaining eigenvectors, after the eigenvalues are obtained. | |
| Real | LanczosCpuTime |
| CPU time for the whole Lanczos procedure, including convergence check and getting eigenvectors (if requested). More... | |
| Real | maxEigenvalueError |
| Maximum eigenvalue error. More... | |
| Real | maxRelativeEigenvalueError |
| Maximum relative eigenvalue error. More... | |
| mkIndex | memoryForLanczosInBytes |
| Memory (in bytes) required for the Lanczos iterations. | |
| bool | allEigenvaluesCheckedConverged |
| allEigenvaluesCheckedConverged is true if all eigenvalues are checked converged. More... | |
| mkIndex | reorthIterCount |
| Number of iterations where reorthogonalization is performed. | |
| mkIndex | reorthVectorCount |
| Total number of vectors against which the reorthogonalization is performed. | |
| Real | reorthVectorRate |
| The value is reorthVectorCount divided by the number of vectors reorthogonalized by full reorthogonalization without double reorthogonalization. More... | |
| mkIndex | localReorthIterCount |
| Number of iterations where reorthogonalization is doubled. | |
| mkIndex | doubleReorthIterCount |
| Number of iterations where local reorthogonalization is performed. | |
| Real | maxOrthLevelRatio |
| Maximum ratio of orthogonality level. More... | |
| Real | minOrthLevelRatio |
| Minimum ratio of orthogonality level. More... | |
Detailed Description
LanczosEigenSolver() returns an instance of this class which gives the information of the Lanczos procedure.
Member Data Documentation
◆ allEigenvaluesCheckedConverged
| bool LanczosInfo::allEigenvaluesCheckedConverged |
allEigenvaluesCheckedConverged is true if all eigenvalues are checked converged.
Otherwise, allEigenvaluesCheckedConverged is false, which means that the maximum number of Lanczos iterations is reached.
◆ LanczosCpuTime
| Real LanczosInfo::LanczosCpuTime |
CPU time for the whole Lanczos procedure, including convergence check and getting eigenvectors (if requested).
- Remarks
- This is the total CPU time for computing the eigenvalues / eigenvectors, excluding reading the data matrix and reporting the result.
◆ maxEigenvalueError
| Real LanczosInfo::maxEigenvalueError |
Maximum eigenvalue error.
The computation requires the eigenvectors, so this parameter is computed only if LanczosOptions::wantEigVec==true as the input of LanczosEigenSolver().
◆ maxOrthLevelRatio
| Real LanczosInfo::maxOrthLevelRatio |
Maximum ratio of orthogonality level.
Let w0(i,j) be the inner product of the ith and jth Lanczos vectors and w0(j) = max{w0(i,j)|i<j}.
Also let w(j) be the estimated upper bound of w0(j). Then maxOrthLevelRatio = max{ w0(j)/w(j) | 1<=j<=numIter }.
- Remarks
- This parameter is computed only if the partial reorthogonalization is applied (i.e. LanczosOptions::reorth==1 as input of LanczosEigenSolver()).
◆ maxRelativeEigenvalueError
| Real LanczosInfo::maxRelativeEigenvalueError |
Maximum relative eigenvalue error.
The computation requires the eigenvectors, so this parameter is computed only if LanczosOptions::wantEigVec==true as the input of LanczosEigenSolver().
◆ minOrthLevelRatio
| Real LanczosInfo::minOrthLevelRatio |
Minimum ratio of orthogonality level.
Let w0(i,j) be the inner product of the ith and jth Lanczos vectors and w0(j) = max{w0(i,j)|i<j}.
Also let w(j) be the estimated upper bound of w0(j). Then minOrthLevelRatio = min{ w0(j)/w(j) | 1<=j<=numIter }.
- Remarks
- This parameter is computed only if the partial reorthogonalization is applied (i.e. LanczosOptions::reorth==1 as input of LanczosEigenSolver()).
◆ reorthVectorRate
| Real LanczosInfo::reorthVectorRate |
The value is reorthVectorCount divided by the number of vectors reorthogonalized by full reorthogonalization without double reorthogonalization.
- If no reorthogonalization (i.e. LanczosOptions::reorth==0 as input of LanczosEigenSolver()) is performed, then reorthVectorRate=0.0.
- If full reorthogonalization (i.e. LanczosOptions::reorth==2 as input of LanczosEigenSolver()) is performed without double reorthogonalization, then reorthVectorRate=1.0.
- If partial reorthogonalization (i.e. LanczosOptions::reorth==1 as input of LanczosEigenSolver()) is performed, usually 0.0<reorthVectorRate<1.0.
The documentation for this struct was generated from the following file:
