GLOSSARY TERM

What is Learning Rate Decay?

A strategy of reducing the learning rate iteratively during model training.
Learning rate decay allows for large weight updates early in training for rapid progression, then gradually shrinks the step size to ensure the model settles precisely into the global or a robust local minimum without oscillating.

Dynamic Scheduling

Configure complex learning rate schedules effortlessly in TacTech MLOps.