remember_data

remember_data

remember_data()[source]

Behaviour flag: whether the estimator remembers all data seen.

• String name: "remember_data"

• Public behaviour flag

• Values: boolean, True / False

• Example: True

• Default: False (transformers), True (forecasters)

Developer tag that modifies behaviour of the boilerplate layer.

If set to True, the estimator will remember all data seen, and store it internally in the self._X and/or self._y attributes, in pd.DataFrame resp hierarchical DataFrame formats pd-multiindex or pd_multiindex_hier, if hierarchical data is passed.

This behaviour can be useful for models with incremental learning capabilities that require access to all previously seen data in their internal logic. Alternatively, the tag can also be useful for debugging.

If no X or y is passed to the estimator, self._X and self._y will be set to None.

Calls to update will update the data stored in the internal attributes as well, using pandas update operations.

If the remember_data tag is set to True, the fit_is_empty tag must be False, even if _fit is empty, because boilerplate writes to self._X in fit.