       Re: Sorting Eigenvalues/EigenVectors

• To: mathgroup at smc.vnet.net
• Subject: [mg8173] Re: Sorting Eigenvalues/EigenVectors
• From: murray at math.umass.edu (Murray Eisenberg)
• Date: Mon, 18 Aug 1997 03:06:56 -0400
• Organization: University of Massachusetts, Amherst
• Sender: owner-wri-mathgroup at wolfram.com

```Joseph C. Slater (jslater at cs.wright.edu) wrote:
: Is there an easy way to sort the results of EigenSystem so that the
: eigenvalues are in increasing order? I can pull them out of the results and
: sort them, but the association of the eigenvalues with the eigenvectors is
: lost. I've been RTFMing for quite a while, and I still can't figure this
: out.

Here's one way -- undoubtedly the most elegant.

a = {{16,19,1},{18,10,20},{10,5,3}};
esys = N[Eigensystem[a]]
evals = esys[]; evect = esys[]
(* Sort[evals] gives the sorted list.  That's not all you want. *)
where = Flatten[Map[Position[Sort[evals], #]&, evals]]
evals[[where]]
evects[[where]]

Now evals[[where]] and evects[[where]] will have corresponding
eigenvalues and eigenspace bases in the same, sorted positions.

Position tells you where a single item in a list is (among other
uses), and then one Map's that to make it work on each element of the
list evals.

One of the design features of Mathematica I don't like is that Sort is
the "basic" function.  I prefer a language like APL, where the
"position-of-in-sorted-order" is the more basic function; from that,
an actual sort just takes idexing.  This seems to be one of the design
principles of APL that Wolfram ignored when he built upon his
knowledge of APL in designing Mathematica.

By the way, the eigenvalues in the example above consist of one real
and two conjugate complex numbers.  The sort order seems to be sort on
real part first, then sort on imaginary part in case of ties.  The
Sort function allows a user-designated sorting function so that you
could, e.g., sort on modulus (in the complex sense of the term), or
even in the case of all real eigenvalues sort in order of decreasing
absolute value -- the most common thing wanted, I suppose.

--
Murray Eisenberg                       Internet:  murray at math.umass.edu
Mathematics & Statistics Dept.            Voice:  413-545-2859 (W)
University of Massachusetts                       413-549-1020 (H)
Amherst, MA 01003                           Fax:  413-545-1801

```

• Prev by Date: Re: Threading objects of unequal length
• Next by Date: Re: ListPlot with little plus signs instead?
• Previous by thread: Re: Sorting Eigenvalues/EigenVectors
• Next by thread: Re: Sorting Eigenvalues/EigenVectors