adaptivesplit package

Subpackages

• adaptivesplit.base package
  • Submodules
  • adaptivesplit.base.learning_curve module
  • adaptivesplit.base.power module
  • adaptivesplit.base.resampling module
  • adaptivesplit.base.split module
  • adaptivesplit.base.utils module
• adaptivesplit.csv_interface package
  • Module contents
• adaptivesplit.sklearn_interface package
  • Submodules
  • adaptivesplit.sklearn_interface.learning_curve module
  • adaptivesplit.sklearn_interface.power module
  • adaptivesplit.sklearn_interface.resampling module
  • adaptivesplit.sklearn_interface.split module
  • adaptivesplit.sklearn_interface.utils module

Submodules

adaptivesplit.config module

class adaptivesplit.config.CustomConfigParser(defaults=None, dict_type=<class 'dict'>, allow_no_value=False, *, delimiters=('=', ':'), comment_prefixes=('#', ';'), inline_comment_prefixes=None, strict=True, empty_lines_in_values=True, default_section='DEFAULT', interpolation=<object object>, converters=<object object>)

Bases: ConfigParser

Custom ConfigParser implementing interpolation. Inherits from configparser.ConfigParser. Used to read AdaptiveSplit’s configuration file which contains the default values for the AdaptiveSplit main class and its ‘stopping_rule’ method. Parameters for the main class are specified in the ‘adaptivesplit.sklearn_interface.split’ section of the documentation, while the parameters for the stopping rule are described here.

Args:

min_training_sample_size (int):

If not 0, the stopping rule considers necessary reaching a number of samples equal to this paramater before finding a stop point. If ‘min_score’ is not equal to -inf then reaching the minimum training sample size and the minimum score acts as precondition for all the other rules to be evaluated. Defaults to the corresponding value in the configuration file.

target_power (float):

Target power to reach before a stopping point can be found. If 0, power rule is not evaluated. Defaults to the corresponding value in the configuration file.

alpha (float):

Target p-val to reach before the slope of the power curve is calculated. Defaults to the corresponding value in the configuration file.

min_score (int or float):

Minimum score to reach before finding a stopping point. See also the ‘min_training_sample_size’ argument for more information. Defaults to the corresponding value in the configuration file.

min_relevant_score (int or float):

If not 0, learning curve rule is evaluated and a stopping point is found when the scores stabilize (useful in case of power curve plateaus). Defaults to the corresponding value in the configuration file.

min_validation_sample_size (int):

If not 0, specifies the minimum size of the validation sample, which coincides with the last possible stopping point. Defaults to the corresponding value in the configuration file.