Editing Transfer Learning (section)

== <span style="color: #FFFFFF;">Remembering</span> ==
* '''Pre-trained model''' — A model trained on a large, general dataset (e.g., ImageNet for vision, Wikipedia for NLP) whose weights serve as a starting point for a new task.
* '''Fine-tuning''' — The process of continuing training of a pre-trained model on task-specific data, adjusting weights to specialize behavior.
* '''Feature extraction''' — Using a pre-trained model as a fixed feature extractor: freeze all its weights and add a new trainable head for the new task.
* '''Domain adaptation''' — Adapting a model trained on one data distribution (source domain) to perform well on a different but related distribution (target domain).
* '''Source domain''' — The original domain on which the model was pre-trained.
* '''Target domain''' — The new domain or task to which you are transferring knowledge.
* '''Domain shift''' — The difference in statistical distribution between source and target domains.
* '''Frozen layers''' — Layers of a pre-trained model whose weights are not updated during fine-tuning.
* '''Trainable layers''' — Layers whose weights are updated during fine-tuning (typically the head and possibly last few blocks).
* '''Head''' — The task-specific output layer(s) added on top of a pre-trained backbone for a new task.
* '''Backbone''' — The main body of a pre-trained model (the feature extractor), as opposed to the task-specific head.
* '''ImageNet''' — A 1.2-million-image classification dataset; models pre-trained on ImageNet are the standard starting point for most computer vision tasks.
* '''BERT''' — A pre-trained transformer encoder; the standard starting point for many NLP fine-tuning tasks.
* '''Domain-adaptive pre-training''' — Additional pre-training on in-domain unlabeled data before task-specific fine-tuning.
* '''Zero-shot transfer''' — Applying a model trained on one task directly to a new task without any task-specific training (e.g., CLIP for zero-shot image classification).
</div>

<div style="background-color: #006400; color: #FFFFFF; padding: 20px; border-radius: 8px; margin-bottom: 15px;">