dqa.tasks.ml.Training#

class dqa.tasks.ml.Training(input_name: str, labels_name: str, algo: Algorithm, weights_name: Optional[str] = None, additional_params: Optional[Dict] = None, direct_params: Optional[Dict] = None, **kwargs)#

Performs training with a machine learning model.

If necessary, the data arrays are concatenated across all machines and measurements in the dataset.

Parameters:
  • input_name (str) – The training data row. Only one value is possible.

  • labels_name (str) – The label data row, also only one value.

  • algo (Algorithm) – An instance of the dqa.configuration.configuration.Algorithm class representing the algorithm.

  • weights_name (str, default=None) – If set, then the corresponding data row is taken as sample weights.

  • additional_params (dict, default=None) – Dictionary with key value pairs to be passed to the model. The values are strings, which are replaced by the corresponding data rows.

  • direct_params (dict, default=None) – Parameters that will be passed directly to the fit function of the model.

Methods

finish()

Can perform actions that are required to clean up after the task has finished, e.g.

in_out_default

input_output_dataset

input_output_machine

input_output_mode

input_output_name

log

modify_data_row

modify_dataset

modify_dataset_dict

modify_machine

modify_measurement

perform_training

set_logging_level

transfer_metadata