This is the code repository accompanied the manuscript titled "Leveraging the Hankel norm approximation and data-driven algorithms in reduced-order modeling." It implements the two-stage algorithm described in the paper by combing the regularized block-AAA algorithm and the modified HNA algorithm. The block-AAA part was heavily adapted from the code in the block_AAA repository. The benchmark examples were taken from the SLICOT benchmark collection. The code was implemented using MATLAB 2023b. In addition, the Control System Toolbox is needed.
Every MATLAB file starting with the word Driver is used to reproduce an experiment in the manuscript. The functions block_aaa and block_aaa_disk contain implementation of the block-AAA part. The function modified_HNA contains the implementation of the modified HNA part. Every other MATLAB file is a helper function.
Every experiment can be reproduced using a MATLAB script.
- To reproduce the experiment in section 4.4, run
Driver_stability.m. - To reproduce the experiment in section 5.4, run
Driver_toy_hilbert.m. - To reproduce the experiment in section 7.1, run
Driver_modified_HNA.m. - To reproduce the experiment in section 7.2, run
Driver_eady.m. - To reproduce the experiment in section 7.3, run
Driver_CD_player.m.
The scripts Driver_eady.m and Driver_CD_player.m employs the block-AAA algorithm with a regularizer. To disable the regularizer, set the regularizing parameter to zero.