Any Plans For Log Space For Hyperparameter Support (Log Argument)? This Is Supported Config Space

Answered

any plans for log space for hyperparameter support (log argument)? this is supported config space https://github.com/automl/ConfigSpace/blob/772665a86bca0d5b0c58ec7e7ea42bf5daadcbd3/ConfigSpace/hyperparameters.pyx

Happy to have a look at this :)

cdef class UniformFloatHyperparameter(FloatHyperparameter): def __init__(self, name: str, lower: Union[int, float], upper: Union[int, float], default_value: Union[int, float, None] = None, q: Union[int, float, None] = None, log: bool = False, meta: Optional[Dict] = None) -> None: """

  				
Posted 
	3 years ago

					More
				  		
  Report
		
					RobustFox47
				
					0
					 × 1

Votes Newest

Answers 3

Hi RobustRat47
What do you mean by "log space for hyperparameter" , what would be the difference ? (Notice that on the graph itself you can switch to log scale when viewing in the UI) ?
Or are you referring to the hyper parameter optimization, allowing you to add log space ?

  				
Posted 
	3 years ago

					More
				  		
  Report
		
					AgitatedDove14
				
					0
					 × 1

lmk if I can expand on this more 🙂

  				
Posted 
	3 years ago

					More
				  		
  Report
		
					RobustFox47
				
					0
					 × 1

Hi AgitatedDove14 ,

I noticed that ClearML parses clearml.automation.UniformParameterRange to configuration space to be used with BOHB. When I've used BOHB previously I can use UniformFloatHyperparameter from the configuration space package that allows me to set a parameter in logspace. That is the range is defended by something like numpy.logspace rather than numpy.linspace

  				
Posted 
	3 years ago

					More
				  		
  Report
		
					RobustFox47
				
					0
					 × 1

Write your answer

987 Views

3 Answers

3 years ago

one year ago