automl / SMAC3

SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimization
https://automl.github.io/SMAC3/v2.1.0/
Other
1.04k stars 220 forks source link

Unclarity in the Documentation #1085

Open LukasFehring opened 7 months ago

LukasFehring commented 7 months ago

Description

When working with smac there are two issues I struggled with:

  1. Some classes do not contain Class Docstrings, while others do. For example AlgorithmConfiguraitonFacade and BlackBoxFacade and RunHistoryLogScaledEncoder, do not have them. While some of them are basic in principle, this makes it significantly harder to understand what is the difference between different components
  2. I did not find an explanation for the default optimization output (Maybe I missed it?). In order to understand the runhistory file, I had to find out the exact place of logging and check what the dict that is written as `json' looks like. Especially the data part
    "data": [
    [
      1,
      null,
      662124363,
      null,
      0.06746031746031744,
      0.05632376670837402,
      1,
      1700216610.3126414,
      1700216610.3698032,
      {}
    ],

    is not self-explanatory. Additionally the data is a list and configs is a dictionary with keys, increasing the difficulty of matching a configuration to the evaluated cost, since the configuration with key 3 has to be accessed as data[2].

  3. In the description of abstract components AbstractIntensifier, a link to Advanced Usage would help.
  4. Advanced Usage should be navigated by a menu (also makes links for 3 more useful).

PS: The issue template implies that one should give a label to issues, but I one does not have the necessary privileges.

alexandertornede commented 7 months ago

Thanks for posting this!