Defines interface for all ACA algorithms and implements classical ACA. More...
#include <TLowRankApx.hh>
Public Member Functions | |
| virtual TMatrix * | build (const TBlockCluster *cl, const TTruncAcc &acc) const |
| virtual TMatrix * | build (const TBlockIndexSet &block_is, const TTruncAcc &acc) const |
| virtual bool | has_statistics () const |
| indicate if algorithm provides statistics | |
Detailed Description
template<typename T>
class HLIB::TACA< T >
Adaptive cross approximation (ACA) is a heuristic for computing a low rank approximation of a given dense matrix by successively removing specific pairs of rows and columns (crosses) from the matrix until the rest is below some threshold (defined by block-wise accuracy).
Due to the algorithm, only the matrix coefficients in form of a TCoeffFn are needed, permitting the straightforward adaption of existing implementations for the construction of H-matrices.
The costs are linear in the dimension of the block and quadratic in the rank.
Member Function Documentation
◆ build() [1/2]
|
virtual |
build low rank matrix for block cluster bct with rank defined by accuracy acc
Implements TLowRankApx.
◆ build() [2/2]
|
virtual |
build low rank matrix for block index set block_is with rank defined by accuracy acc
