Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Start visualisation #19

Merged

Conversation

benedikt-voelkel
Copy link
Collaborator

  • add first own visualisation functionality
    (not using optuna's for now because that cannot handle a large number
    of parameters and makes all plots useless because they become too
    crowded)

  • plot functionality in Inspector

    • importance plot
    • parallel coordinates
    • parameters vs. loss value (per trial)
    • parameter correlations
    • pair-wise scatter plots (basically more in-depth corelation)
  • added plot showing history of steps, hits (per detector), loss to cut
    tuning evaluation

  • slight update of README

  • add a few short one/two-character variable names to .pylintrc

@benedikt-voelkel
Copy link
Collaborator Author

benedikt-voelkel commented Jul 15, 2022

Task-agnostic plots
These should work independent of the specifics of the underlying specifics of the optimisation task

importance_parameters
slices
parallel_coordinates
parameter_correlations

pairwise_scatter

This plot is specific for the cut tuning example
steps_hits_history

@benedikt-voelkel
Copy link
Collaborator Author

benedikt-voelkel commented Jul 15, 2022

Hej @mconcas
I am opening this PR for discussion so far (draft).
See above possible visualisations. I am sure, the optics of some of the plots can be enhanced and I will do so for some of them.
In addition, the code can be optimised, some plots might be compiled in a bit of a brute-force way?!

But most importantly, I think we should get in some visualisation for evaluation and interpretation purposes as soon as possible to enhance the evaluation of future optimisation tasks.

One thing: This brings in pandas and seaborn. Up to discussion whether or not we want to use them or if we want to implement our own implementations where I am currently using them.

(and just seeing now that the CI is run, sorry, I will update the requirements if we agree on them)

@mconcas
Copy link
Collaborator

mconcas commented Jul 18, 2022

Thanks for opening this.

But most importantly, I think we should get in some visualisation for evaluation and interpretation purposes as soon as possible to enhance the evaluation of future optimisation tasks.

Agree, and indeed in general, as you are also the first "customer" I'd say whatever makes sense for your convenience should be added, for the moment. We'll figure out eventually if something is superfluous.

One thing: This brings in pandas and seaborn

That is fine to me, see above comment.

@benedikt-voelkel
Copy link
Collaborator Author

Ok, sounds good!
So I will tidy it up and make it ready to be merged!

* add first own visualisation functionality
  (not using optuna's for now because that cannot handle a large number
   of parameters and makes all plots useless because they become too
   crowded)

* plot functionality in Inspector
  * importance plot
  * parallel coordinates
  * parameters vs. loss value (per trial)
  * parameter correlations
  * pair-wise scatter plots (basically more in-depth corelation)

* added plot showing history of steps, hits (per detector), loss to cut
  tuning evaluation

* slight update of README

* add a few short one/two-character variable names to .pylintrc

* add pandas and seaborn as deps
@benedikt-voelkel benedikt-voelkel marked this pull request as ready for review July 18, 2022 09:23
@benedikt-voelkel benedikt-voelkel merged commit 53b1b24 into AliceO2Group:master Jul 18, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants