Tune

class nimble.Tune(values=None, start=None, end=None, change=None, changeType=None)

Triggers hyperparameter optimization to occur during training.

Provide or generate the possible argument values to use during the hyperparameter optimization process (defined by a Tuning object) and the best argument will be used to train the learner. A list of predetermined values can be passed as the values parameter or a range of values can be constructed using the start, end, change and changeType parameters. Only end is required in this case, the other parameters will be assigned default values if not explicitly set (see below).

Parameters:

  • values (list) – A list of argument values to use for tuning. Either this or end must not be None.

  • start (int, float) – The inclusive start value of the values in a range. Default to 0 and only applies when end is not None.

  • end (int, float) – The inclusive end value of the values in a range. Either this or values must not be None.

  • change (int, float) – The amount by which to changeType the data in the range. The changeType parameter will dictate whether this uses addition or multiplication. Defaults to 1 and only applies when when end is not None.

  • changeType (str) – Either ‘add’ or ‘multiply’ to indicate how the change will be used to generate the range. Defaults to ‘add’ and only applies when end is not None.

See also

Tuning, nimble.train

Keywords

cross-validation, parameter, argument, hyperparameters, tuning, optimization, cross validate, learn, hyper parameters, hyperparameters, choose, grid search, GridSearchCV