Glossary term
Glossary term
Evaluation and Benchmarks
The mathematical formula or metric that a model aims to optimize. For example, the objective function for linear regression is usually Mean Squared Loss. Therefore, when training a linear regression model, training aims to minimize Mean Squared Loss.
In some cases, the goal is to maximize the objective function. For example, if the objective function is accuracy, the goal is to maximize accuracy.
See also loss.
Created for this library
An ML team picks an objective function that aligns with the business metric so optimization improvements translate to product wins.
A risk modeling team selects an objective function with calibration in mind so downstream decisions can use predicted probabilities directly.
A research team chooses an objective function based on the desired error tolerance, picking quantile loss when tails matter more than the mean.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License