SAS Tutorial | Transfer Learning

In this SAS How To Tutorial, Robert explains transfer learning, a relatively new area of machine learning that leverages information extracted from one set of distributions. Transfer learning refers to a situation where information extracted from one setting (referred to as the source) is exploited in another setting (referred to as the target) to improve the learner’s performance. Because biases can cause negative transfer learning (meaning the learner’s performance degrades because of transfer learning), machine learning practitioners should be cautious when using transfer learning.

Transfer learning biases include frequency and context biases.  Although, there are strategies that can be leveraged to mitigate the impact of these biases on your model’s performance.

Chapters
00:00 – Welcome
00:52 – What is frequency bias?
02:27 – What is context bias?
03:43 – Transfer learning in an unsupervised model
05:11 – Transfer learning in a supervised model
05:58 – What is an adaptation layer?

See Robert’s other videos on deep learning

Paper for reference
Weiss, K., T. Khoshgoftaar, and D. Wang. 2016. “A survey of Transfer Learning.” Journal of Big Data 3: Article number: 9

Learn more about SAS Software

SUBSCRIBE TO THE SAS USERS YOUTUBE CHANNEL #SASUsers #LearnSAS

ABOUT SAS
SAS is a trusted analytics powerhouse for organizations seeking immediate value from their data. A deep bench of analytics solutions and broad industry knowledge keep our customers coming back and feeling confident. With SAS®, you can discover insights from your data and make sense of it all. Identify what’s working and fix what isn’t. Make more intelligent decisions. And drive relevant change.

CONNECT WITH SAS