# Transfer Learning
Status: public
Confidence: medium (0.78) (verified)
Last verified: 2026-05-28
Generation: ai_structured

## TL;DR
Transfer learning reuses knowledge learned from one task, dataset, or domain to improve learning on another. In modern deep learning, this often means pretraining on a large corpus or dataset and then adapting the model to a smaller target task.

## Core Explanation
Common strategies include feature extraction, linear probing, full fine-tuning, domain adaptation, and parameter-efficient tuning. Transfer can help when labeled data is scarce, but it can also fail when the source and target differ too much.

## Detailed Analysis
Transfer learning appears in computer vision, NLP, speech, and multimodal models. The evidence should name the source task, target task, and adaptation method because "transfer" can mean anything from frozen feature reuse to full model fine-tuning.

## Further Reading
- Pan and Yang on transfer learning
- Yosinski et al. on feature transferability
- BERT

## Related Articles

- [Adversarial Machine Learning: Attacks, Defenses, and Robustness Engineering](../adversarial-machine-learning.md)
- [AI for Drug Repurposing: Identifying New Uses for Existing Drugs Through Machine Learning](../ai-drug-repurposing.md)
- [AI for Employee Experience: Onboarding, Learning, and Internal Communications](../ai-employee-experience.md)