[Feature] Changing b-values

[oliverchampion]

Feature description

So for many algorithms, we now allow to give b-values at initialisation and at fit-time. This is confusing and probably is not supported for all algorithms

Describe the solution

I think it is cleaner to only allow to give b-values at initialisation, as some algorithms then initialise e.g. priors, and changing b-values may rise issues. On the other side, for some specific applications, this may require re-initialising the algorithm unnecessarily often.

Alternatively, we can support changing b-values at fit-time , but we would have to add unit tests to see whether this works well.

Describe alternatives

No response

Additional context

No response

Are you working on this?

None

[Devguru-codes]

Hi @oliverchampion sir, I would like to work on this!

I think b-values only at initialisation is the cleaner approach, some algorithms like IAR_LU_biexp build internal structures (e.g., dipy gradient_table) from b-values at init, so changing them at fit-time would require re-initialization and could introduce subtle bugs.

I can audit all algorithms, standardize the interface in OsipiBase, and add unit tests. Please assign me. Thank you.

[oliverchampion]

I agree that it seems the most straightforward. Not sure it needs unit testing, we just remove values as optional input, and add an error message if it is given. @IvanARashid and @etpeterson, do you agree with this solution?

[IvanARashid]

I agree that it seems the most straightforward. Not sure it needs unit testing, we just remove values as optional input, and add an error message if it is given. @IvanARashid and @etpeterson, do you agree with this solution?

Agreed

[Devguru-codes]

Ok, I will create a plan and then get it validated. After that, I will start working. Thank you,