learning rate

Go back to the [[AI Glossary]]

A scalar used to train a model via gradient descent. During each iteration, the gradient descent algorithm multiplies the learning rate by the gradient. The resulting product is called the gradient step.

Learning rate is a key hyperparameter.