Tasks

• Tasks are a set of parameters that define an experiment.

• Autoregression, Interpolation, and Fault Detection all

share the same fields: procedure, name and metrics. - Name is used to identify is the task is autoregression, interpolation, or fault detection. - Metrics are used to identify the means of evaluating performance in a specific task.

• Must be part of gridds.tools.metrics.py.

• Procedure is used to define if we are doing fit then predict or fit_transform.

• Autoregression tasks have parameters that define delay` and horizon.

  • delay: number of measurements to be used in history for autoregression.

  • horizon: number of measurements to be predicted in future for autoregression.

• Here are some examples:

  default_autoregression = {
      'name': 'autoregression',
      'procedure': ['fit_transform'],
      'metrics': [mae, rmse],
      'delay': 5,
      'horizon': 1
  }
  
  long_autoregression = {
      'name': 'autoregression',
      'procedure': ['fit_transform'],
      'metrics': [mae, rmse],
      'delay': 20,
      'horizon': 5
  }
  
  
  default_impute = {
      'name': 'impute',
      'procedure': ['fit_transform'],
      'metrics': [mae, rmse]
  }
  
  default_fault_pred = {
      'name': 'fault_pred',
      'procedure': ['fit','predict'],
      'metrics': [binary_crossentropy]
  }
  
  unsupervised_fault_pred = {
      'name': 'fault_pred',
      'procedure': ['predict'],
      'metrics': [binary_crossentropy]
  }