Build automatic reporting for classification algorithm comparison

[lkorczowski]

Build automatic reporting for classification algorithm comparison

• [x] first and foremost: have a proper and clean bruxism_detection notebook ! see #19
• [x] build script to applied our methods to all data (the notebook should include the output) see #22
• [ ] automatically export insights and build a report, e.g. :
• [ ] a. (unsupervised) visualization of N bruxism detected events for visual inspection (assessing false positive), N being low enough to be easy to assess manually and high enough to be statistically valid. This method should be dropped as soon as we have .c @RobinGuillard will do that during Sprint 3 #28
• [x] b. (unsupervised) temporal distribution of the events (this metric will be used for subsequent sleep state analysis), see #22
• [ ] c. (supervised) AUC score and other metrics (but we need tagging) discussed in #26

After that, any additional method (e.g. electrode disconnection) should be benchmarked using this report. It will save us time and sanity.

Discussed but not in the scope of this issue:

• problem when person is awake, see #24

Originally posted by @lkorczowski in https://github.com/lkorczowski/Tinnitus-n-Sleep/issues/16#issuecomment-605898163

[lkorczowski]

not sure it is done yert

[lkorczowski]

@RobinGuillard : I think should put your remarks here on the metrics we need for the report ( also the caveat of detecting awake in discussed in #24 )

[lkorczowski]

Closing as this issue is split with

• automatic benchmarking #26
• NB-related tasks in SPRINT 3 #28