MathGroup Archive 2012

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

Search the Archive

Eigensystem bug in Mathematica 7.0.1 on Windows 7 (64 bit) for

  • To: mathgroup at smc.vnet.net
  • Subject: [mg127949] Eigensystem bug in Mathematica 7.0.1 on Windows 7 (64 bit) for
  • From: jdbates <batej6 at gmail.com>
  • Date: Tue, 4 Sep 2012 05:48:26 -0400 (EDT)
  • Delivered-to: l-mathgroup@mail-archive0.wolfram.com
  • Delivered-to: l-mathgroup@wolfram.com
  • Delivered-to: mathgroup-newout@smc.vnet.net
  • Delivered-to: mathgroup-newsend@smc.vnet.net

Hello all,

I seem to have encounter a bug in the Eigensystem function in Mathematica 7.0.1 for Windows 7 (64 bit).

I'm attempting to solve for the eigenvectors of a covariance function on with support on a random lattice.  The process works up to 80 points in the lattice, but as soon as I include that 81st point the Eigensystem function fails (it generates spuriously negative eigenvalues and incorrect eigenvectors).

Here's an example code demonstrating this behavior:

**********************************

Cov[x_]=Exp[-x^2/a^2];

numx=81;
d=1;
a0=0.1;

xs=Table[RandomReal[{-d,d}],{n1,1,numx}]

CM=Table[Cov[xs[[n1]]-xs[[n2]]]/.{a->a0},{n1,1,numx},{n2,1,numx}];

{el,ev} = Eigensystem[CM]

**********************************

For numx <= 80, the Eigensystem command works fine.  For numx > 80, it fails.  This behavior seems to be independent of choice of lattice points (recomputing the xs numerous times did not change this behavior).  It also seems relatively independent of the width parameter a0, although I've found that if I make a0 sufficiently small such that the covariance matrix becomes sparse then the Eigensystem function seems to work correctly.

After searching this board, I've encountered a couple of threads regarding similar issues with Eigensystem for previous versions of Mathematica, but none for 7.0.1 or later.  Is this a known issue, and if so is there a workaround?



  • Prev by Date: Re: split the sublists into parts according to some rules
  • Next by Date: Re: DSolve for a real function
  • Previous by thread: Missing simplification for ArcSin
  • Next by thread: Re: Eigensystem bug in Mathematica 7.0.1 on Windows 7 (64 bit) for