Lampisan opened a new pull request #1489: URL: https://github.com/apache/systemds/pull/1489
[SYSTEMDS-3167] Alternative Implementations of Eigen Decomposition This commit adds implementations of the following matrix algorithms in LibCommonsMath: Lanczos algorithm, QR decomposition, QR algorithm and Householder transformation (unused). Furthermore tests for the Lanczos and the QR algorithm are added. Lanczos algorithm is numerically unstable and therefore sometimes produces incorrect results. QR decomposition is working as intended. QR algorithm works for symmetric matrices. While it should theoretically also work for non-symmetric matrices, it currently does not produce all correct eigenvectors in some cases. Furthermore a runtime-exception is thrown if the matrix has complex eigenvalues or the algorithm hasn't converged. We are under the impression that our implementations of these algorithms are really slow compared to the existing implementation. Although we only tested this with a single thread this implementation is orders of magnitudes slower. Co-authored-by: Nikolaus Grogger <[email protected]> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
