James Kent

### inverse_transforms that do ensure 4D output: (uses `signal_extraction` in `inverse_transform`) - Parcellations - NiftiLabelsMasker ### inverse_transforms that do not ensure 4D output: (uses `unmask` in `inverse_transform` directly or indirectly)...

@NicolasGensollen @bthirion WDYT? For enforcement of 4D outputs: - @bthirion: https://github.com/nilearn/nilearn/pull/2725#issuecomment-794399854 - @GaelVaroquaux (2015): https://github.com/nilearn/nilearn/issues/460#issuecomment-100767188 Against enforcement of 4D outputs: - The people that make use of this 1D input/3D...

If you are tagged, this means you've worked on a repository that uses `inverse_transform` from nilearn. The question we are trying to answer is: **If one passes a 1D input...

some relevant discussions from scikit-learn: - scikit-learn/scikit-learn#4509 - scikit-learn/scikit-learn#4511 - scikit-learn/scikit-learn#4252 Moving this conversation from just outputs to inputs and outputs, I see three options: - transform(4D) -> 2D output;...

It's going to be a pain to get this to work with python 3.6 (maybe not possible without a refactor of what libraries we are using), I could see the...

readthedocs build error is unrelated to this pull request, and is related to readthedocs build process #798, will see if a fix gets merged in later today to rebuild the...

LGTM, thanks for all your hard work on this @yifan0330! this is an excellent contribution!

Also thanks to @tsalo and @adelavega for testing and offering reviews and having heated (but friendly) conversations about code structure.

great work @tsalo!

this has been completed with the nimad module in NiMARE