Fine-Tuning

Published:

Fine-tuning is the process of taking a pre-trained model and adjusting it so it performs well on a new, more specific task. Instead of training a model from the ground up, you start with one that already learned broad, useful features from a large dataset. From there, you adapt it by training some or all of its layers on your new dataset.

A typical workflow looks like this: load the pre-trained model, replace the final layer with one designed for your new output, and begin training. Often the early layers are frozen at first so the model keeps the general knowledge it already has, while the new layers learn the specifics of your task. Later, you can unfreeze more layers and fine-tune them carefully with a lower learning rate to avoid overwriting what the model previously learned.

Follow us on Facebook and LinkedIn to keep abreast of our latest news and articles

Other definitions

All Definitionmore-content
Fairness Auditingmore-content
Federated Learningmore-content
Forecastingmore-content
Framework Creationmore-content
Feature Engineeringmore-content