Short answers, pulled from the story.
Transfer learning is a machine learning technique where knowledge gained from training on one task is reused to improve performance on a related task. For example, a model trained to recognize cars can apply that knowledge when learning to recognize trucks.
Bozinovski and Fulgosi published the first paper addressing transfer learning in neural network training in 1976, providing a mathematical and geometrical model of the concept.
At his NIPS 2016 tutorial, Andrew Ng said that transfer learning would become the next major driver of commercial success in machine learning, following supervised learning.
Yes. In the 2020 paper "Rethinking Pre-Training and self-training," Zoph and colleagues reported that pre-training can hurt accuracy in some conditions and advocated for self-training instead.
Transfer learning has been applied to cancer subtype discovery, building utilization, general game playing, text classification, digit recognition, medical imaging, and spam filtering.
A 2020 study found that due to their similar physical natures, transfer learning is possible between electromyographic (EMG) muscle signals and electroencephalographic (EEG) brainwave signals, and the relationship works in both directions.