Сравнение методов

	Дообученная (fine-tuned) свёрточная нейронная сеть	Дообученная рекуррентная нейронная сеть
Область	Глубокое обучение	Глубокое обучение
Семейство	Machine learning	Machine learning
Год появления	2012–2014	2015–2018
Автор метода	Yosinski, J. et al. (theoretical basis); practice widespread from Krizhevsky et al. 2012 onward	Popularised by Howard & Ruder (ULMFiT, 2018); RNN fine-tuning concept developed iteratively in the NLP community from ~2015
Тип	Transfer learning technique (supervised fine-tuning)	Transfer learning / sequential model adaptation
Основополагающий источник	Yosinski, J., Clune, J., Bengio, Y., & Lipson, H. (2014). How transferable are features in deep neural networks? Advances in Neural Information Processing Systems, 27.	Howard, J. & Ruder, S. (2018). Universal Language Model Fine-Tuning for Text Classification. Proceedings of ACL 2018, 328–339.
Другие названия	Fine-tuned CNN, CNN fine-tuning, CNN transfer learning with fine-tuning, adapted convolutional network	Fine-Tuned RNN, RNN Fine-Tuning, domain-adapted RNN, pre-trained RNN with downstream adaptation
Связанные	5	6
Сводка	Fine-tuning a CNN means starting from a network already trained on a large dataset — typically ImageNet — and continuing training on a smaller target dataset so the model adapts its learned visual features to a new task. This approach dramatically reduces the data and compute required to reach strong performance compared with training from scratch.	A Fine-Tuned Recurrent Neural Network (RNN) starts from a model pre-trained on large corpora or time-series data and adapts its weights to a specific downstream task through controlled gradient updates. The approach dramatically cuts the labeled data needed for strong sequence modeling performance in text classification, named entity recognition, sentiment analysis, and related tasks.