mne-bids-pipeline
mne-bids-pipeline copied to clipboard
mne-tools
Any thoughts about nipype?
This seems like an interesting and potentially useful project.
I'm curious if there was any thought to making a nipype (https://nipype.readthedocs.io/en/latest/) type interfaces?
I don't use nipype very often but I have found it useful in the past to standardize some large batch processing. Just a thought.
And maybe https://neuropycon.github.io/ephypype/ to make it easy to use nipype for these tasks?
It would be great if you can try this on CTF data to identify potential pain points in using this. I suggested the same to @AdoNunes
as for neuropycon, give it a try but maybe joblibs does the job already and you don't need caching.
what would be the benefit of nipype? our current approach is pure python so no need to mix languages and packages. I find our approach simple in understanding what we pass as input and get as output
What I think nipype has to offer is scalability, flexibility and sharing, but that comes at the expense of some of the simplicity and clarity.
Some of the things I think are nice about nipype are its
- workflow visualization/diagrams (see the middle bottom of https://miykael.github.io/nipype_tutorial/notebooks/example_preprocessing.html)
- workflow hotstarts in the event of errors etc.
- optional overwriting etc. controlled at the workflow level.
- sharability: if the nodes here were in nipype I could just use them.
- its a robust pipelining tool - why invent another one?
Also it seems like someone is supporting the mne-c and mne-pthon function calls (https://github.com/nipy/nipype/issues/3032)
Like i said before I don't really use nipype but it is something that i think i would use if the barrier to entry was alittle bit lower.
I stay open but I feel this is premature here. We have something that works and that needs to be consolidated. Once we have a stable code/solution why not looking into different strategies.