Skip to content

Transfer Learning

Transfer learning is a technique where a model trained on one task is adapted to improve performance on a different but related task. In practice, it often builds on supervised or self-supervised pretraining.


  1. Train (or use) a model on a source task with abundant data
  2. Transfer the learned features/weights to a new target task
  3. Fine-tune on the target task’s data (usually smaller)

TypeDescriptionExample
Feature ExtractionUse a pre-trained model as a fixed feature extractorUse ResNet features for a custom image classifier
Fine-TuningRetrain some or all layers on the new taskFine-tune BERT for sentiment analysis
Domain AdaptationAdapt a model to a related domainImageNet model → medical imaging

BenefitDescription
Less data requiredLeverage pre-trained knowledge; need fewer labeled examples
Faster trainingStart from learned weights, not random initialization
Better performancePre-trained features often generalize well

  • NLP: Fine-tune GPT or BERT for summarization, Q&A, or classification
  • Computer Vision: Use ImageNet-trained models for medical imaging or satellite imagery
  • Speech: Adapt a general ASR model to a specific accent or domain vocabulary

If the source and target tasks are too different, transfer can hurt performance (called negative transfer). Always validate on the target task.

Exam Tip: When you want to leverage knowledge from a pre-trained model to improve performance on a new, related task with less data and compute, transfer learning is the most suitable approach.