# Transfer learning and the low-label regime

Source: https://onco.cc/terms/transfer-learning/  
OnCo record `transfer-learning` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Transfer learning reuses what a model learned on one task or dataset to do better on a related task with few labels, the situation for almost every cancer outcome.

## Summary

Transfer learning re-uses knowledge learned from a task to boost performance on a related task (Wikipedia). Its promise in oncology is the low-label regime: outcome labels number in the hundreds, so a model pretrained on unlabelled data should need fewer of them. The gain is largest with fifty to a few hundred labels and tends to vanish as labels grow, and it can be washed out entirely by widening the feature set, so claims should state the label count.

## Fields

- Kind: Term
- Last checked: 2026-09-24
- Also known as: transfer learning; low-label regime; few labels; label-efficient; data-efficient learning; few-shot fine-tuning
- Tags: cansim-terms

## Notes

- Listed in the CanSim terms map 1.0.0 (docs/onco/terms.json, generated 2026-09-24), CC BY 4.0, attribution: CanSim project, an open, public-data-first cancer foundation-model programme; CanSim page path /terms/transfer-learning.

## Sources

- Wikipedia: https://en.wikipedia.org/wiki/Transfer_learning
- Wikipedia: https://en.wikipedia.org/wiki/Transfer_learning

## Connected records

- terms: [Cancer AI vocabulary (CanSim terms map)](https://onco.cc/terms/cancer-ai-vocabulary/), [Fine-tuning versus frozen features, and LoRA](https://onco.cc/terms/fine-tuning-vs-frozen/), [Self-supervised pretraining (SSL)](https://onco.cc/terms/self-supervised-pretraining/), [Zero-shot prediction](https://onco.cc/terms/zero-shot/)

---
JSON: https://onco.cc/api/v1/entities/transfer-learning.json