End goal: get previous semester's web service up and running, compatable with our new data organization, and new preprocessing functions. Also, do some general 'spring cleaning' and separate our research repo (jupyter notebooks and such) from the production pipeline, also document all code that we know will be used in the final.
[x] create a separate repository folder organization for our 'production' code, separate from 'research' code.
[x] integrate the benchmarking and pipelining code, so that benchmark metrics are reported during all pipeline runs.
[x] run through all the code put into this new repo, delete everything that isn't actually used. re-factor for efficiency and use the auto-doc strings for everything that is used.
[ ] set up continuous integration and put hooks into github
[ ] set up the web service from last semester, and integrate these things into it, making modifications when necessary. Final goal is the old web service running with our new preprocessing tasks, with benchmarking metrics reported at the end (along with the usual plots).
[ ] (bonus) see if the benchmarking utilities can be extended to other data kinds (fMRI, anything in BIDS format) and extend the auto-bench utilities to handle this kind of data.
Researchy Stuff
End goal: Decision of which interpolation method (including possible new ones) should be implemented in the pipeline. Also, decision of which order it should be applied (e.g. before or after noise reduction?)
[ ] See how spherical interpolation interacts with the denoising techniques. E.g. which should go first? For now, interpolate the channels that Nicolas did. Do a big comparison (like noise reduction) for raw, spherical splines interpolated, inverse distance weighting interpolated, and nicolas preprocessed. Do this befre and after the noise reduction, then cross compare the two.
[x] Look at wavelet coefficients for 'bad' signals and 'good' signals. What if we just did something like inverse distance weighting, but on the coefficients? This would interpolate signals using both a spatial and temporal component. If this looks promising, methods.md for it (and add it to the benchmarking stuff).
Deliverable
Notes
Engineeringy Stuff
End goal: get previous semester's web service up and running, compatable with our new data organization, and new preprocessing functions. Also, do some general 'spring cleaning' and separate our research repo (jupyter notebooks and such) from the production pipeline, also document all code that we know will be used in the final.
repositoryfolder organization for our 'production' code, separate from 'research' code.Researchy Stuff
End goal: Decision of which interpolation method (including possible new ones) should be implemented in the pipeline. Also, decision of which order it should be applied (e.g. before or after noise reduction?)