Skip to main content

Timeline for answer to NSolve gives additional solutions that don't satisfy the equations! by Daniel Lichtblau

Current License: CC BY-SA 3.0

Post Revisions

8 events
when toggle format what by license comment
Nov 1, 2012 at 16:40 comment added dbm @Licthblau, Excellent! Thanks a lot for your detailed explanation. In short, we need to be extra careful while using NSolve for systems with approximate coefficients. I already knew discriminant variety and homotopy continuation method. But I didn't know about the near discriminant variety issue, and specially that it could affect solving systems so drastically. Thanks a lot again! dbm368
Nov 1, 2012 at 14:19 comment added Daniel Lichtblau Edited to add explanation.
Nov 1, 2012 at 14:19 history edited Daniel Lichtblau CC BY-SA 3.0
Explained arisal of extraneous large solutions
Nov 1, 2012 at 0:17 history edited J. M.'s missing motivation CC BY-SA 3.0
deleted 1 characters in body
Oct 31, 2012 at 23:57 comment added dbm Thanks for your comment and your request to others to stop down-voting. The problem comes from particle theory where I have a nonlinear parametric system of equations and I want to find all real solutions of it. In particular, I want to see if the no. of real solutions is constant for the parametric system. The present system is at a particular parameter-point (a generic point). I started having doubt about the infinite solutions was from my cross-checking the solutions with homotopy continuation method which gave only two solutions always instead of 8 by NSolve. Thanks again.
Oct 31, 2012 at 23:39 comment added Daniel Lichtblau I'll edit my response tomorrow to address that issue of "infinite" solutions. Meanwhile you might mention where the system came from or the nature of the problem. That might help when I add an explanation.
Oct 31, 2012 at 23:04 comment added dbm That worked! However, if I still think there is more fundamental problem here: NSolve gives solutions at infinity. If you see the actual solutions in sol, some variables take values of the order 10^16. These are solutions at infinity for a reasonable numerical precision say 10^13 or so. So there is a catch22 here. If we choose WorkingPrecision->50, the solutions at infinity are considered as affine solutions, if we choose the default WorkingPrecision then these solutions still appear but don't satisfy the equations! May be I am explaining it not properly?!
Oct 31, 2012 at 22:51 history answered Daniel Lichtblau CC BY-SA 3.0