From help-request at octave dot org Thu Dec 23 09:34:07 2004 Subject: invfreq: computation using LS-TLS approach From: "Pascal A. Dupuis" To: help at octave dot org Date: Thu, 23 Dec 2004 08:45:16 -0600 Hello, invfreq solve the problem of fitting some parametrised frequency response to some given data. The main step are: - construct a design matrix (R) and a target vector (P) - find the set of parameters (Theta) which best maps the design matrix to the target vector The actual code uses LS, meaning only P is supposed to be noised. But some columns of R are constructed from the target frequency response, and are thus also noised. Furthermore, if the frequency is noised, the whole R matrix is noised. I have coded an alternative to solving Theta=R\P using QR, where the noised part is handled by TLS, and the unnoised part by standard LS. I have performed a few simulations, and it appears that this mixed approach gives parameters which are more accurate than with ordinary LS. Before submitting patches, I would like to ask to other octave users what should be the best way to pass one supplemental argument to invfreqs and invfreqz, namely the resolution method. I imagine a string with possible values: 'LS' (default), 'TLS', or 'LS-TLS'. TIA Pascal Dupuis -- Dr. ir. Pascal Dupuis K. U. Leuven, ESAT/ELECTA (formerly ELEN): http://www.esat.kuleuven.ac.be/ Kasteelpark Arenberg, 10; B-3001 Leuven-Heverlee, Belgium Tel. +32-16-32 10 21 -- Fax +32-16-32 19 85 ------------------------------------------------------------- Octave is freely available under the terms of the GNU GPL. Octave's home on the web: http://www.octave.org How to fund new projects: http://www.octave.org/funding.html Subscription information: http://www.octave.org/archive.html -------------------------------------------------------------