NeuroDataDesign / orange-panda-f16s17

Automated EEG data analysis.

5 stars 5 forks source link

December 5th #42

Closed rmarren1 closed 7 years ago

rmarren1 commented 7 years ago

Commit

http://cortex.jhu.edu:8011

Status	Task	Deliverable
🐼 :	Scale Robust PCA for webservice and Algorithms.md for Removing Outliers	demo of the format of the code results in prep report @tsunwong123
:panda_face:	Flow Chart	available in the about tab on web service (http://cortex.jhu.edu:8011) @rmarren1
🐼 :	Choose which pipeline processes to use from EEGLAB set	python notebook with simulations @nkumarcc
:panda_face:	Algorithm, Simulation, and Analysis for KS Artifact Rejection Method (Algorithm.md)	@mnatenz1
:panda_face:	Simplify steps for demo (user doesn't choose, we choose)	details of changes @rmarren1
:panda_face:	Document EVERYTHING	details of docs created here @rmarren1

Reach

Status	Task	Deliverable
⏳ :	Use discriminability to decide best threshold for bad electrode detection	python notebook with simulations @nkumarcc

jovo commented 7 years ago

[ ] documentation should all go in webpage, possibly in "about", at least linked to from webpage, so there is a coherent place to get everything
[ ] how was algorithms.md?
[ ] what is our current recommendation for bad electrode detection?
[ ] did KS test "work"?
[ ] median of what? let's incorporate into pipeline!
[ ] let's find out how people define bad ICA components?

ebridge2 commented 7 years ago

@nkumarcc, discussed in class, but when using multi part algorithms (ie, discriminability is a distance step, a rdf step, and then a discr step) make your stuff fit the first step, and leave the rest the same. Ie, only touch the distance step and make that work for you; dont toucg working implementations of rdf and discriminability. Gives you less to rewrite, and concentrates down the number of sources of errors if you make a mistake adjusting code.