Re: Kalman Filter &Smoother, Subspace Identification
- To: mathgroup at smc.vnet.net
- Subject: [mg37667] Re: [mg37643] Kalman Filter &Smoother, Subspace Identification
- From: "Johannes Ludsteck" <johannes.ludsteck at wiwi.uni-regensburg.de>
- Date: Fri, 8 Nov 2002 02:14:15 -0500 (EST)
- Organization: Universitaet Regensburg
- Sender: owner-wri-mathgroup at wolfram.com
Dear Roy, I have used the TimeSeries Kalman Filter functions about 3 Years ago, but not extensively. The functions are multivariate and support abstract definition of the state space system (i.e. very general nonlinear structures of the model). I cannot tell you whether it is fast since my problems were rather small. Of course, it is possible to conduct likelihood estimation, but I cannot tell whether this is fast. You have to use FindMinimum... Of course, the code is available, since TimeSeries is simply a large Package. However, it may be complex and you probably need at least intermediate Mathematica experience to understand the stuff. I don't know what you mean with subspace identification. Best regards, Johannes Ludsteck On 7 Nov 2002 at 6:41, Roy Mendelssohn wrote: > I know this question arises fronm time to time, but has anyone > implemented a modern, multivariate Kalman filter and smoother? By > modern, I mean there have been a lot of improvements to the algorithm in > the last few years that both spedd up and complement the computations. > > On a related matter, does anyone know anything about the Time Series > package form Wolfram, and especially it's Kalman algorithm (my inquries > to Wolfram have produced a fact sheet and not much else, and the fact > sheet doesn't tell you very much ). Is the algorithm univariate of > multivariate, is it readily tied in some how with an optimization > algorithm so that it can be used for likelihood estimation, is the code > avaiable if you purchase it sothat it can be modified? > > And finally, has anyone implemented subspace identification algorithms, > such as those in the System identification Toolbox? > > TIA, > > Roy Mendelssohn > > > <><><><><><><><><><><><><><><><><><> Johannes Ludsteck Institut fuer Volkswirtschaftslehre Lehrstuhl Prof. Dr. Moeller Universitaet Regensburg Universitaetsstrasse 31 93053 Regensburg Tel +49/0941/943-2741