Skip to content

Joint Optimization

Some algorithms support joint optimization of the task across multiple outputs, and can improve results compared to using multiple independent models. To leverage multiple outputs in PostgresML, you'll need to substitue the standard usage of pgml.train and pgml.predict with pgml.train_joint and pgml.predict_joint. The _joint functions are identical, except train_joint takes an array of y_column_names TEXT[], and predict_joint returns an array of outputs correspondingly.

Read more at scikit-learn.