skrub.Expr.skb.freeze_after_fit

Expr.skb.freeze_after_fit()

Freeze the result during pipeline fitting.

Note this is an advanced functionality, and the need for it is usually an indication that we need to define a custom scikit-learn transformer that we can use with .skb.apply().

When we use freeze_after_fit(), the result of the expression is computed during fit(), and then reused (not recomputed) during transform() or predict().

Returns:

The expression whose value does not change after fit()

Examples

>>> import skrub
>>> X_df = skrub.toy_orders().X
>>> X_df
   ID product  quantity        date
0   1     pen         2  2020-04-03
1   2     cup         3  2020-04-04
2   3     cup         5  2020-04-04
3   4   spoon         1  2020-04-05
>>> n_products = skrub.X()['product'].nunique()
>>> transformer = n_products.skb.get_pipeline()
>>> transformer.fit_transform({'X': X_df})
3

If we take only the first 2 rows nunique() (a stateless function) returns 2:

>>> transformer.transform({'X': X_df.iloc[:2]})
2

If instead of recomputing it we want the number of products to be remembered during fit and reused during transform:

>>> n_products = skrub.X()['product'].nunique().skb.freeze_after_fit()
>>> transformer = n_products.skb.get_pipeline()
>>> transformer.fit_transform({'X': X_df})
3
>>> transformer.transform({'X': X_df.iloc[:2]})
3