I've been doing some research into computer vision lately and constantly come across the ability of residual networks to improve performance. Intuitively I think I grasp them, however, I struggle to understand why they are not more prolific in other non-image based fields. I'm currently specialized in NLP and I rarely see them.

My question is, what non-image use-cases have you seen residual networks succeed? And why do you think they did?


1 Answer 1


Residual Networks are designed to avoid the vanishing gradient problem in deep neural networks. LSTM cells, which are commonly used in NLP, have a 'natural' way to avoid these, see e.g. (How does LSTM prevent the vanishing gradient problem?).

However there is research using residual connections combined with LSTMs for NLP, see Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.