visualizations and storage for experiments

[cmry]

This is a big one, which requires the set-up of a database back-end (done already) and a (web) front-end:

• [x] Dump serialized config files to JSON database.
• [x] Represent the performance in a way that general metrics can be shown.
  • [x] project
  • [x] name
  • [x] training_set (i.a.)
  • [x] testing_set (i.a) -- this and above probably need to be abstracted from loaders
  • [x] string representation of the used features
  • [x] string NAME of the classifier used
  • [x] ~~POS / NEG f1-scores (could be put in a graph)~~ micro f-1
• [ ] Able to overview and compare experiments visually.
  • [x] Flat Performance bar.
  • [x] Plotting performance on data proportions.
  • [ ] Summary of experiment configurations.
  • [x] Confusion matrices.
  • [ ] Aggregate performances in one report.
  • [ ] t-SNE?
• [ ] Insight into feature importances.
  • [x] LIME evaluation.
  • [ ] Coof. representations.

[cmry]

First rough work on the front-end done in https://github.com/cmry/omesa/commit/705ce59006b38eb94e098cc452ecad80feaf792e. For now, first there needs to be some representation in the database that makes relevant calls a bit easier. Some of those high level representations should be:

• project
• name
• training_set (i.a.)
• testing_set (i.a) -- this and above probably need to be abstracted from loaders
• string representation of the used features
• string NAME of the classifier used
• POS / NEG f1-scores (could be put in a graph)

[cmry]

LIME evaluation (from this paper) has been added since https://github.com/cmry/omesa/commit/3da48276999e1b401d52889a774c2437ea15aec9. Currently for binary classification only!