AMAN ARORA: “Demystifying Transfer Learning - what's exactly going on and why is it so popular?”
Автор: DataScienceSydney
Загружено: 2019-06-17
Просмотров: 202
Transfer Learning is perhaps the reason why deep learning is so successful and is able to achieve great results across various domains. In the words of an expert data scientist and fast.ai founder Jeremy Howard 'Every time we put transfer learning into anything, we generally make it much better'. In mid-2018 Jeremy Howard and Sebastian Ruder put transfer learning to test in NLP and the proposed method(ULMFiT) significantly outperformed existing state-of-art on six text classification tasks reducing the error by 18-24% on the majority of datasets. Through this talk let's demystify transfer learning and understand what's really going on under the hood and why is this method so successful.
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке:
