Adding new documentation example for Blackbox Optimization #1073
Conversation
| ############################################################################ | ||
| # Designing hyperparameter space to optimize | ||
| # ****************************************** | ||
| # Obtaining the hyperparameter domain for the SVM model to tune. We use the |
There was a problem hiding this comment.
I still think we should just specify the upper and lower bounds based on some tutorial as that'll simplify the example quite a bit. How about following the LibSVM FAQ?
There was a problem hiding this comment.
I couldn't find a range for these HPs we can refer to and use. The FAQs for LIBSVM and other related resources generally point to a good value to work with.
How about referring to FABOLAS and using their (e ^ -10, e ^ 10) range for these 2 HPs?
There was a problem hiding this comment.
Ha, it's not in the FAQ, it's in the guide (https://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf).
There was a problem hiding this comment.
Ah, and could you please use the 2**-5 notation instead of giving the full floating point value?
Co-authored-by: Matthias Feurer <feurerm@informatik.uni-freiburg.de>
| general = dict( | ||
| name="sklearn_bbo_example_flow", | ||
| description=("Running BO on SVM using OpenML"), | ||
| external_version="bayesian-optimization==1.2.0", |
There was a problem hiding this comment.
Could you please add scikit-learn here, also can you use the installed version of bayesian-optimization as a dependency?
There was a problem hiding this comment.
use the installed version of bayesian-optimization as a dependency?
I didn't quite understand. v.1.2.0 is the one that I have locally to create the example.
There was a problem hiding this comment.
What I meant is that you query the version via bayes_opt.__version__ instead of putting a fixed version into the script.
There was a problem hiding this comment.
alright, understood and makes sense now
but how to handle this when version or __version__ is not available from the package?
added a couple lines of code to do this
Co-authored-by: Matthias Feurer <feurerm@informatik.uni-freiburg.de>
|
We still want this, in particular an example custom flows with trace iterations. |
What does this PR implement/fix? Explain your changes.
A tutorial that guides users to use OpenML for BBO.