Sam Coveney

Thanks @RafaelNH I think this might be related to the test you are doing having no error on the data - can I confirm that this is true? I thought...

@RafaelNH when you get the chance, can you confirm whether the test you performed was on noiseless data please?

Thanks @RafaelNH I have added tests for DTI, will get them done for DKI and try to address what to do when fitting noiseless signals (issue is that the noise...

Have updated dki tests as well, should all be done. @RafaelNH I note that you didn't get an error, but just a warning, is this okay given the user error...

> Warnings: we switched to zero warnings policy on the CI's and I can see some warnings popping up during tests. Can you address them? I must have missed this,...

Note that the code format checks often fail while insisting on things that would fail pep8... shame. Also: > [Line 90:24](https://github.com/dipy/dipy/blob/b3fef2e95f3a020b0121ea846b011c791174d7aa/dipy/reconst/weights_method.py#L90): [E712](https://duckduckgo.com/?q=pep8%20E712) comparison to False should be 'if cond is...

> > we can not 😅 , you can do like here: > > https://github.com/dipy/dipy/blob/60669b888e7bc322a8b973f3a0b8121e4fd7f7d1/dipy/reconst/dti.py#L2037 Hmm does this do what is needed? Surely that would need argwhere, for starters? Anyway,...

Looks like `cls_fit_dki` is randomly failing on noisy data with valid minimum signal. This is true whether or not using my new code here... I will look into it next...

@skoudoro @RafaelNH I believe I have done everything possible (and the requested changes are out of date), can we please look at getting this pulled if no issues? I can...

Since not pulled yet, I made a small modification for the definition of weights, which allows to provide a custom function to define outliers (which then sets the corresponding weights...