Glossary term
Glossary term
Foundations
A loss function that calculates the absolute value of the difference between actual label values and the values that a model predicts. For example, here's the calculation of L1 loss for a batch of five examples:
L1 loss is less sensitive to outliers than L2 loss.
The Mean Absolute Error is the average L1 loss per example.
Click the icon to see the formal math.
See Linear regression: Loss in Machine Learning Crash Course for more information.
Created for this library
A demand forecasting team uses L1 loss because business reviewers focus on median absolute error rather than squared error.
A risk team uses L1 loss for robustness in a model where occasional extreme labels would dominate squared-loss training.
A research team uses L1 loss when training a model on time-series data with heavy-tailed targets.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License