MathGroup Archive 2011

[Date Index] [Thread Index] [Author Index]

Search the Archive


  • To: mathgroup at
  • Subject: [mg119718] SingularValueDecomposition
  • From: John Snyder <jsnyder at>
  • Date: Sat, 18 Jun 2011 06:15:18 -0400 (EDT)

I read Jon McLoone's recent post on the WolframBlog concerning the 
solution of the drunken sailor's walk problem by using a Markov chain 
transition probability matrix. He mentions that it may also be possible 
to solve the problem using the SingularValueDecomposition function, but 
he does not illustrate this. I am trying to figure out how this could be 

Here is a simple "toy" example. Assume that I have the following Markov 
chain transition probability matrix m where each row sums to 1:


Assuming that I start in the position 2 (column 2 out of 3, in the first 
of 4 rows) I want to find the so-called "fixed point", the ultimate 
state density function, as the number of steps goes to infinity. I know 
that I can do this numerically using MatrixPower as follows (here is 100 
steps which appears to be more than enough in this case):

In[19]:= MatrixPower[N[m],100][[2]]//Chop
Out[19]= {0,0,0,0,0,0,0,0,0,0,0,0,0.809663,0.190337}

I believe that it is also possible to get this same result by using the 
SingularValueDecomposition function, but I cannot figure out how to get 
this to work. Can someone please show me how to use 
SingularValueDecomposition to get the same answer to this question? I 
know there are other ways to solve this, but I am really interested in 
using SingularValueDecomposition in this case. Thanks.

  • Prev by Date: Re: Filtering of the higher frequency terms.
  • Next by Date: Re: capacitor equation solution not working
  • Previous by thread: Re: Filtering of the higher frequency terms.
  • Next by thread: Re: SingularValueDecomposition