IVIM VarPro implementation throws infeasible 'x0'

[ShreyasFadnavis]

Description

This bug occurs rarely when the number of iterations for Differential Evolution is low.

Will follow-up with a PR with:

• Default
• Warning when user reduces the maxiter below a certain value.

#1826 will also help fix some of these issues.

Way to reproduce

-- Cannot be reproduced with a fixed recipe.

[ShreyasFadnavis]

@Garyfallidis @arokem Please feel free to let me know any other approach you have in mind. Another alternative I have is to switch to LM implementation is the solution is infeasible (may result in a linear fit though). I think, after Cythonizing the implementation, this bug would go away anyways!

[ShreyasFadnavis]

#1839 fixes this. Closing!

[yinting527]

@ShreyasFadnavis Hello, I'm trying out ivim varpro reconstruction using dipy, and getting error message " x0 is infeasible." while not setting the bounds, and "Inconsistent shapes between bounds and x0." after setting bounds. the 'trr' method won't give this error. May I ask how should I solve this? Thanks!

[ShreyasFadnavis]

@ShreyasFadnavis Hello, I'm trying out ivim varpro reconstruction using dipy, and getting error message " x0 is infeasible." while not setting the bounds, and "Inconsistent shapes between bounds and x0." after setting bounds. the 'trr' method won't give this error. May I ask how should I solve this? Thanks!

Hi @xxhdd ! Thank you for reaching out and reporting this. As for the bounds, there are certain things to keep in mind when using VarPro as it's a slightly different optimizer as compared to trr. Can I take a quick look at your code? You can mail it to me on shreyasfadnavis@gmail.com

On a side note, are you using a mask?
I am also making a new update for the code with a speeded up version and will handle bounds better!

[arokem]

@xxhdd : You can post the code here, so that people other than @ShreyasFadnavis can see it, and so that we don't have to wonder what happened a year down the line, when we're dealing with any other issues.

[ShreyasFadnavis]

PR #1993 should fix this. The issue was that @xxhdd is working with animal data and not the brain. the x0 is not feasible error has been handled by reusing the same procedure used in trr.

[skoudoro]

Hi @ShreyasFadnavis,

What is the status of this issue? I do not find the related PR, Can you point it to me?

Thank you

[ShreyasFadnavis]

Hi @skoudoro ! This issue has been resolved for @xxhdd . However I do intend to make a new PR with an updated version of this... I will also add the error handling withing that 👍

[skoudoro]

ok, so I let this issue open and look forward to seeing your PR @ShreyasFadnavis!

[seanyang9527]

Hi, I've encountered the same error message. Despite spending days going through the PRs, I couldn't find a solution.