Transfer Learning Explained for Experts


Transfer learning refers to the method in machine learning where knowledge from a source task is transferred to a target task to improve learning performance.

This often occurs through the use of pretrained models, whose parameters serve as a starting point for training on the target task.


The challenge lies in finding suitable representations that are generalizable and avoid overfitting.

Transfer learning encompasses various scenarios, including cross-domain learning, multi-task learning, and few-shot learning.


Definition:
“Transfer learning aims to improve learning in a new task through the transfer of knowledge from a related task that has already been learned.”


Source:
Pan, S. J., & Yang, Q. (2010). A Survey on Transfer Learning. IEEE Transactions on Knowledge and Data Engineering, 22(10), 1345-1359.