10

In Why does this stiff BVP return unevaluated but a similar IVP produce a solution?, I addressed a somewhat common problem with NDSolve, namely that when the shooting method fails due to stiffness or a singularity in one of the trial solutions, the method fails with giving the user any clue except that stiffness or a singularity was encountered (Most recently, it happened in Solving a steady-state viscous Burger's equation with NDSolve.) It does not say how or where the stiffness arises, and often it is due to a poor choice of initial conditions by the "Shooting" method.

1) I meant the post to address this specific problem about the "Shooting" method. The recent, linked question in parentheses above has two questions. So I hope it is not considered a duplicate, because I wanted a narrow question focused on a somewhat common problem with differential equations that I've seen on the site.

2) I think it's a principally because it leaves the user unnecessarily in the dark. I'd like to seek consensus that it be tagged a bug. I've reported it, but WRI may or may not agree. Technically, one might say it is merely a suggested improvement. But I've seen several people waste their time on the problem without a clue how as to why NDSolve is failing or where to look for options and alternatives that might solve it. For instance, the problem is not pointed out in Numerical Solution of Boundary Value Problems; nor does the tutorial point out the possibility of using ParametricNDSolve to manually search for a solution.

My main point is to ask whether we should tag it a . The secondary point is to explain why I posted the Q&A, which might seem to some to be a dupe.

| |
  • 11
    Whether or not it is a bug, the poor performance of NDSolve certainly reflects badly on Wolfram, Inc. Although I do not use Maple, I surmise from posts on StackExchange by others that Maple solves ODE boundary value problems much better than Mathematica does. For that matter, DSolve also performs poorly on many seemingly straightforward problems. As @Mr.Wizard once remarked, Wolfram, Inc should stop adding new features until it gets the existing features to work properly. – bbgodfrey Nov 6 '19 at 1:56

You must log in to answer this question.

Browse other questions tagged .