ConfigSpace
ConfigSpace copied to clipboard
automl
Domain specific language for configuration spaces in Python/Cython. Useful for hyperparameter optimization and algorithm configuration.
This is the solution I came up with after opening issue #336. I have also noted that I could not use `Integer()` with a default value because the integer default...
As a automl-developer I want to to create a network with n layers of different with. Currently I cannot sample the with easily, I have to create an extra entry...
Similar to https://github.com/automl/ConfigSpace/issues/73 . However, if the order of two conjunctions (maybe also forbiddenclauses) is different, then the two configurations will be considered different ones: ``` from ConfigSpace import ConfigurationSpace,...
[Well-known precision issue](https://floating-point-gui.de/errors/comparison/) Would you be able to use `np.isclose` or something similar for the comparison? I think the problem came from [here](https://github.com/automl/ConfigSpace/blob/main/ConfigSpace/hyperparameters/uniform_float.pyx#L104-L110). ```python In [1]: import ConfigSpace as CS...
The following code crashes, because `NormalFloatHyperparameter` attempts to set a default value without respecting the bounds. ``` cs = ConfigurationSpace( name="myspace", space={ "a": Float("a", bounds=(1, 2), distribution=Normal(1, 0), log=True), },...
Version: 0.7.2 https://automl.github.io/ConfigSpace/main/guide.html this is stated on the website: conf = cs.sample_configuration() conf['max_iter'] = 42 print(conf['max_iter']) 42 but it returns an error - I would have expected it to be...
https://github.com/automl/ConfigSpace/blob/671465e2899f7b435fa23970d872c30177014f3e/ConfigSpace/read_and_write/pcs_new.py#L481-L488
To decouple the code for bounded and unbounded normal hyperparameters.
https://github.com/automl/ConfigSpace/blob/88051ebb7c34b6c8e69777c75dbcbc120b7720fd/ConfigSpace/hyperparameters.pyx#L2865C2-L2900
Reference PR #282 Both `_compute_normalization` and `get_max_density` will compute the pdf over the entire integer range of `NormalIntegerHyperparameter` and `BetaIntegerHyperparameter`. This is usually fine unless the ranges get massive, i.e....