Simultaneously improving the efficiency and compression of the adaptive cross approximation algorithm

Abstract

This paper proposes an adaptation of the conventional ACA algorithm for block-wise impedance matrix compression, yielding a substantial speed-up as well as a higher compression rate. The recently published stochastic Frobenius norm estimation is complemented with a new approach to allow applying it to blocks that represent touching regions. The stochastic method is then combined with SVD recompression to simultaneously eliminate the stochastic uncertainty and optimize the compression rate.