Share via

RandomParameterSampling Class

Defines random sampling over a hyperparameter search space.

Initialize RandomParameterSampling.

Constructor

RandomParameterSampling(parameter_space, properties=None)

Parameters

Name Description
parameter_space
Required
dict

A dictionary containing each parameter and its distribution. The dictionary key is the name of the parameter.
properties
dict

A dictionary with additional properties for the algorithm.

Default value: None
parameter_space
Required
dict

A dictionary containing each parameter and its distribution. The dictionary key is the name of the parameter.
properties
Required
dict

A dictionary with additional properties for the algorithm.

Remarks

In this sampling algorithm, parameter values are chosen from a set of discrete values or a distribution over a continuous range. Examples of functions you can use include: choice, uniform, loguniform, normal, and lognormal. For example,


   {
       "init_lr": uniform(0.0005, 0.005),
       "hidden_size": choice(0, 100, 120, 140, 180)
   }

This will define a search space with two parameters, init_lr and hidden_size. The init_lr can have a uniform distribution with 0.0005 as a minimum value and 0.005 as a maximum value, and the hidden_size will be a choice of [80, 100, 120, 140, 180].

For more information about using RandomParameter sampling, see the tutorial Tune hyperparameters for your model.

Attributes

SAMPLING_NAME 

SAMPLING_NAME = 'RANDOM'