What is layer freezing in transfer learning and when should it be used?

Layer freezing prevents pretrained layers from updating during training.

Freezing is used when pretrained features are general and only task-specific layers need training. Later, selective unfreezing allows fine-tuning for better adaptation.