Re: [eigen] unit-tests |

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]

*To*: eigen@xxxxxxxxxxxxxxxxxxx*Subject*: Re: [eigen] unit-tests*From*: Benoît Jacob <jacob@xxxxxxxxxxxxxxx>*Date*: Wed, 24 Oct 2007 12:04:04 +0200

Hi Andre, I did survive fortunately :) but I am now under very high work pressure. I have currently a lot of teaching to do at my university, I'll give a talk in 2 weeks, and I must prepare my postdoc applications. Whether or not I'll find a nice postdoc position depends on whether I'll do good research in the next monthes, starting now, and these present weeks are especially important as I'll come to visit in November the university at which I'd like the most to do a postdoc, so I should work as much as possible until then. So I don't currently have much time for coding. It'll get better in January, as I will be done with this heavy teaching duty. Anyway, we already have most of the basics there. The Core module is mostly there, though I have a few changes in mind. What we need now is other modules: - Geom, doing low-dimensional things such as rotations/quaternions, euclidean stuff such as Gram-Schmidt, and projective geometry as MatrixP did in Eigen1. - LU, doing all things involving LU decomposition, including inverse, determinant, and a solver of systems of linear equations. If we had LU and Geom modules, and quickly implemented regression on top of LUas we did in Eigen1, that would already be a great milestone as that would mean that Eigen2 could do all what Eigen1 did (and much more). So at this stage it would become possible to port existing apps from Eigen1 to Eigen2. Other modules that we need, and that had no counterpart in Eigen1: - QR, doing all things involving QR decomposition, that is diagonalization and spectrum computation (eigenvalues). This would allow to compute the spectral radius, the operator norm, etc. As an application we could easily propose a solver for polynomial equations P(x)=0. (by doing a QR on the companion matrix of P). - SVD, the Singular Values Decomposition, and all what it would allow to do. Ultimately, regression should use that instead of LU. Cheers, Benoit On Wednesday 24 October 2007 10:42:35 Andre Krause wrote: > hy benoit, did you survive your Disputation / Rigorosum ? > i hope you are not completely burned out and can continue to work on > eigen2. > > cheers, andre

**Attachment:
signature.asc**

**References**:**[eigen] unit-tests***From:*Benoît Jacob

**Re: [eigen] unit-tests***From:*Andre Krause

**Messages sorted by:**[ date | thread ]- Prev by Date:
**Re: [eigen] unit-tests** - Previous by thread:
**Re: [eigen] unit-tests**

Mail converted by MHonArc 2.6.19+ | http://listengine.tuxfamily.org/ |