Deep learning has been making headlines around the world for its unparalleled performance in everything from recognizing cat videos to playing go. However deep learning remains well beyond the reach of average developers and researchers due to its massive compute and data requirements. Transfer learning is an old field of study going through a renaissance that promises to change by re-using portions of deep network models to drastically reduce both compute and data requirements. Come learn about some traditional transfer learning techniques such as word2vec and newer technologies like GPT and BERT.
