Glossary term
Glossary term
Foundations
During training or testing, a mathematical function that calculates the loss on a batch of examples. A loss function returns a lower loss for models that makes good predictions than for models that make bad predictions.
The goal of training is typically to minimize the loss that a loss function returns.
Many different kinds of loss functions exist. Pick the appropriate loss function for the kind of model you are building. For example:
L2 loss (or Mean Squared Error) is the loss function for linear regression.
Log Loss is the loss function for logistic regression.
Created for this library
A risk modeling team selects log loss as its production loss function because it aligns with the calibrated probability the downstream system uses.
A demand forecasting team uses a quantile loss function so the model can produce calibrated demand percentiles for inventory planning.
A research team evaluates several loss functions before picking the one that aligns with the business metric.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License