help on tomsym
in [Matlab]
|
From: Manthos Vogiatzoglou on 2 Mar 2010 18:15 I am interested in calculating analytical gradient and hessian of some multivariate likelihoods. To do this, matrix calculus knowledge is essential, especially if you want to work with the arbitrary n - dimensional case. Untill yesterday I was not aware of any software implementing actuall symbolic matrix calculus therefore I tried to do all the calculations by hand. However, I am not sure that my calculations are actually correct. Today I came across with tomlab and especially tomsym that claims it can calculate analytical derivatives (the symbolic toolbox cannot calculate symbolic matrix calculus derivatives, for more info: Magnus, Jan R.; Neudecker, Heinz (1999 (1988)). Matrix Differential Calculus. Wiley Series in Probability and Statistics (revised ed.). Wiley. pp. 171–173., or in http://ht.econ.kobe-u.ac.jp/~tanizaki/workshop/2007/20070309.pdf). The problem, after the very first day, is that I couldn't find any detailed users guide on tomsym therefore I have some fundamental questions, like, for example: 1. Sometimes the symbolic expression of the derivative is given in a form unfamiliar to me, because it uses some tomlab functions I am not accustomed with, like lookup. Is there a way I can alter the form of the output to standard matlab syntax or at least to another friendlier form? 2. Some other times, instead of a "tomsym" expression I get as a result like that the output is a 4x3 symmat object, without giving me any expession. Is there a command I can use so to "vizualize" the symbolic expression? The tomsym engine seems very prommising but I need some documentation to fully understand its abilities (or I need some of you to answer to all my questions!!!!)! Can anyone tell me where I could find such documentation, if available? Thanks in advance Manthos Vogiatzoglou
From: Wayne King on 2 Mar 2010 19:29 "Manthos Vogiatzoglou" <vogia.m(a)gmail.com> wrote in message <hmk65q$4j8$1(a)fred.mathworks.com>... > I am interested in calculating analytical gradient and hessian of some multivariate likelihoods. To do this, matrix calculus knowledge is essential, especially if you want to work with the arbitrary n - dimensional case. Untill yesterday I was not aware of any software implementing actuall symbolic matrix calculus therefore I tried to do all the calculations by hand. However, I am not sure that my calculations are actually correct. Today I came across with tomlab and especially tomsym that claims it can calculate analytical derivatives (the symbolic toolbox cannot calculate symbolic matrix calculus derivatives, for more info: Magnus, Jan R.; Neudecker, Heinz (1999 (1988)). Matrix Differential Calculus. Wiley Series in Probability and Statistics (revised ed.). Wiley. pp. 171–173., or in http://ht.econ.kobe-u.ac.jp/~tanizaki/workshop/2007/20070309.pdf). The problem, after the very > first day, is that I couldn't find any detailed users guide on tomsym therefore I have some fundamental questions, like, for example: > 1. Sometimes the symbolic expression of the derivative is given in a form unfamiliar to me, because it uses some tomlab functions I am not accustomed with, like lookup. Is there a way I can alter the form of the output to standard matlab syntax or at least to another friendlier form? > 2. Some other times, instead of a "tomsym" expression I get as a result like that the output is a 4x3 symmat object, without giving me any expession. Is there a command I can use so to "vizualize" the symbolic expression? > The tomsym engine seems very prommising but I need some documentation to fully understand its abilities (or I need some of you to answer to all my questions!!!!)! Can anyone tell me where I could find such documentation, if available? > > Thanks in advance > > Manthos Vogiatzoglou Hi Manthos, have you checked their website? They appear to have plenty of documentation. http://tomopt.com/docs/models/tomlab_models.php Wayne
|
Pages: 1 Prev: error when saving: unable to write to MAT-file, File may be Next: modified bessel functions |