Glossary term
Glossary term
Foundations
A type of machine learning algorithm that improves the performance of a model by combining the predictions of multiple models and using those predictions to make a single prediction. As a result, a loss aggregator can reduce the variance of the predictions and improve the accuracy of the predictions.
Created for this library
A multi-task NLP team uses a loss aggregator to combine task losses with tunable weights during shared training.
A research team uses a loss aggregator with dynamic weights so auxiliary losses fade as the model approaches convergence on the main task.
An ML platform team builds a reusable loss aggregator so multi-objective training jobs share a consistent configuration interface.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License