This contribution proposes an adapted version of the popular Adaptive Cross Approximation algorithm for block wise compression of Method of Moments impedance matrices.
The estimation of the matrix block Frobenius norm, necessary to
know the relative error in the compressed representation takes
up a substantial percentage of the total computation time. In the
adapted version it is replaced by a stochastic process with
negligible cost. The involved uncertainly is eliminated using an a
posteriori SVD recompression which, as a fortunate side-effect,
yields an important additional reduction of the compressed rank
CitacióHeldring, A., Ubeda, E., Rius, J. Adapted ACA algorithm with improved efficiency and compression rate. A: Iberian Meeting on Computational Electromagnetics. "X EIEC: X Iberian Meeting on Computational Electromagnetics: 5-8 May 2015, Baeza (Spain)". Baeza, Jaén: 2015.